diff --git a/README.md b/README.md index 054592100bfb30e57ad015badff944719f6a4455..eb19a55a870f085f1c9ccece612a37d9d4f7e867 100644 --- a/README.md +++ b/README.md @@ -696,7 +696,7 @@ DeepSparkInference将按季度进行版本更新，后续会逐步丰富模型 RetinaFace FP16 Supported - - + Supported INT8 diff --git a/models/cv/classification/convnext_base/ixrt/README.md b/models/cv/classification/convnext_base/ixrt/README.md new file mode 100644 index 0000000000000000000000000000000000000000..62bfb23c5f1280728eea1f0e0ca843441e6ad64f --- /dev/null +++ b/models/cv/classification/convnext_base/ixrt/README.md @@ -0,0 +1,59 @@ +# ConvNeXt Base + +## Description + +The ConvNeXt Base model represents a significant stride in the evolution of convolutional neural networks (CNNs), introduced by researchers at Facebook AI Research (FAIR) and UC Berkeley. It is part of the ConvNeXt family, which challenges the dominance of Vision Transformers (ViTs) in the realm of visual recognition tasks. + +## Setup + +### Install + +```bash +# Install libGL +## CentOS +yum install -y mesa-libGL +## Ubuntu +apt install -y libgl1-mesa-dev + +pip3 install tqdm +pip3 install onnx +pip3 install onnxsim +pip3 install tabulate +pip3 install ppq +pip3 install tqdm +pip3 install cuda-python +``` + +### Download + +Pretrained model: + +Dataset: to download the validation dataset. + +### Model Conversion + +```bash +python3 export.py --weight convnext_base-6075fbad.pth --output convnext_base.onnx +``` + +## Inference + +```bash +export DATASETS_DIR=/Path/to/imagenet_val/ +``` + +### FP16 + +```bash + +# Accuracy +bash scripts/infer_convnext_base_fp16_accuracy.sh +# Performance +bash scripts/infer_convnext_base_fp16_performance.sh +``` + +## Results + +| Model | BatchSize | Precision | FPS | Top-1(%) | Top-5(%) | +| -------------- | --------- | --------- | ------- | -------- | -------- | +| ConvNext Base | 32 | FP16 | 219.817 | 83.70 | 96.73 | diff --git a/models/cv/classification/convnext_base/ixrt/build_engine.py b/models/cv/classification/convnext_base/ixrt/build_engine.py new file mode 100644 index 0000000000000000000000000000000000000000..038c15d5af7a51da6386aa0f994e830178330fb3 --- /dev/null +++ b/models/cv/classification/convnext_base/ixrt/build_engine.py @@ -0,0 +1,52 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import cv2 +import argparse +import numpy as np + +import torch +import tensorrt + +def main(config): + IXRT_LOGGER = tensorrt.Logger(tensorrt.Logger.WARNING) + builder = tensorrt.Builder(IXRT_LOGGER) + EXPLICIT_BATCH = 1 << (int)(tensorrt.NetworkDefinitionCreationFlag.EXPLICIT_BATCH) + network = builder.create_network(EXPLICIT_BATCH) + build_config = builder.create_builder_config() + parser = tensorrt.OnnxParser(network, IXRT_LOGGER) + parser.parse_from_file(config.model) + + precision = tensorrt.BuilderFlag.INT8 if config.precision == "int8" else tensorrt.BuilderFlag.FP16 + build_config.set_flag(precision) + + plan = builder.build_serialized_network(network, build_config) + engine_file_path = config.engine + with open(engine_file_path, "wb") as f: + f.write(plan) + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--model", type=str) + parser.add_argument("--precision", type=str, choices=["float16", "int8", "float32"], default="int8", + help="The precision of datatype") + parser.add_argument("--engine", type=str, default=None) + args = parser.parse_args() + return args + +if __name__ == "__main__": + args = parse_args() + main(args) \ No newline at end of file diff --git a/models/cv/classification/convnext_base/ixrt/calibration_dataset.py b/models/cv/classification/convnext_base/ixrt/calibration_dataset.py new file mode 100644 index 0000000000000000000000000000000000000000..ec931c656abf5b2309dc9938490df46e4e8cdb19 --- /dev/null +++ b/models/cv/classification/convnext_base/ixrt/calibration_dataset.py @@ -0,0 +1,112 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import os + +import torch +import torchvision.datasets +from torch.utils.data import DataLoader +from torchvision import models +from torchvision import transforms as T + + +class CalibrationImageNet(torchvision.datasets.ImageFolder): + def __init__(self, *args, **kwargs): + super(CalibrationImageNet, self).__init__(*args, **kwargs) + img2label_path = os.path.join(self.root, "val_map.txt") + if not os.path.exists(img2label_path): + raise FileNotFoundError(f"Not found label file `{img2label_path}`.") + + self.img2label_map = self.make_img2label_map(img2label_path) + + def make_img2label_map(self, path): + with open(path) as f: + lines = f.readlines() + + img2lable_map = dict() + for line in lines: + line = line.lstrip().rstrip().split("\t") + if len(line) != 2: + continue + img_name, label = line + img_name = img_name.strip() + if img_name in [None, ""]: + continue + label = int(label.strip()) + img2lable_map[img_name] = label + return img2lable_map + + def __getitem__(self, index): + path, target = self.samples[index] + sample = self.loader(path) + if self.transform is not None: + sample = self.transform(sample) + # if self.target_transform is not None: + # target = self.target_transform(target) + img_name = os.path.basename(path) + target = self.img2label_map[img_name] + + return sample, target + + +def create_dataloaders(data_path, num_samples=1024, img_sz=224, batch_size=2, workers=0): + dataset = CalibrationImageNet( + data_path, + transform=T.Compose( + [ + T.Resize(256), + T.CenterCrop(img_sz), + T.ToTensor(), + T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]), + ] + ), + ) + + calibration_dataset = dataset + if num_samples is not None: + calibration_dataset = torch.utils.data.Subset( + dataset, indices=range(num_samples) + ) + + calibration_dataloader = DataLoader( + calibration_dataset, + shuffle=True, + batch_size=batch_size, + drop_last=False, + num_workers=workers, + ) + + verify_dataloader = DataLoader( + dataset, + shuffle=False, + batch_size=batch_size, + drop_last=False, + num_workers=workers, + ) + + return calibration_dataloader, verify_dataloader + + +def getdataloader(dataset_dir, step=20, batch_size=32, workers=2, img_sz=224, total_sample=50000): + num_samples = min(total_sample, step * batch_size) + if step < 0: + num_samples = None + calibration_dataloader, _ = create_dataloaders( + dataset_dir, + img_sz=img_sz, + batch_size=batch_size, + workers=workers, + num_samples=num_samples, + ) + return calibration_dataloader diff --git a/models/cv/classification/convnext_base/ixrt/common.py b/models/cv/classification/convnext_base/ixrt/common.py new file mode 100644 index 0000000000000000000000000000000000000000..fd6a84d85fcc4ede8f64ed0e0a2dd7022527a956 --- /dev/null +++ b/models/cv/classification/convnext_base/ixrt/common.py @@ -0,0 +1,81 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import cv2 +import glob +import torch +import tensorrt +import numpy as np +from cuda import cuda, cudart + +def eval_batch(batch_score, batch_label): + batch_score = torch.tensor(torch.from_numpy(batch_score), dtype=torch.float32) + values, indices = batch_score.topk(5) + top1, top5 = 0, 0 + for idx, label in enumerate(batch_label): + + if label == indices[idx][0]: + top1 += 1 + if label in indices[idx]: + top5 += 1 + return top1, top5 + +def create_engine_context(engine_path, logger): + with open(engine_path, "rb") as f: + runtime = tensorrt.Runtime(logger) + assert runtime + engine = runtime.deserialize_cuda_engine(f.read()) + assert engine + context = engine.create_execution_context() + assert context + + return engine, context + +def get_io_bindings(engine): + # Setup I/O bindings + inputs = [] + outputs = [] + allocations = [] + + for i in range(engine.num_bindings): + is_input = False + if engine.binding_is_input(i): + is_input = True + name = engine.get_binding_name(i) + dtype = engine.get_binding_dtype(i) + shape = engine.get_binding_shape(i) + if is_input: + batch_size = shape[0] + size = np.dtype(tensorrt.nptype(dtype)).itemsize + for s in shape: + size *= s + err, allocation = cudart.cudaMalloc(size) + assert err == cudart.cudaError_t.cudaSuccess + binding = { + "index": i, + "name": name, + "dtype": np.dtype(tensorrt.nptype(dtype)), + "shape": list(shape), + "allocation": allocation, + "nbytes": size, + } + print(f"binding {i}, name : {name} dtype : {np.dtype(tensorrt.nptype(dtype))} shape : {list(shape)}") + allocations.append(allocation) + if engine.binding_is_input(i): + inputs.append(binding) + else: + outputs.append(binding) + return inputs, outputs, allocations \ No newline at end of file diff --git a/models/cv/classification/convnext_base/ixrt/config/CONVNEXT_CONFIG b/models/cv/classification/convnext_base/ixrt/config/CONVNEXT_CONFIG new file mode 100644 index 0000000000000000000000000000000000000000..71bf70ad92da305b87f4cd5283b66a7f5d893f0e --- /dev/null +++ b/models/cv/classification/convnext_base/ixrt/config/CONVNEXT_CONFIG @@ -0,0 +1,33 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +# IMGSIZE : 模型输入hw大小 +# MODEL_NAME : 生成onnx/engine的basename +# ORIGINE_MODEL : 原始onnx文件名称 +IMGSIZE=224 +MODEL_NAME=Convnext_base +ORIGINE_MODEL=convnext_base.onnx + +# QUANT CONFIG (仅PRECISION为int8时生效) + # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] + # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) + # QUANT_STEP : 量化步数 + # QUANT_SEED : 随机种子 保证量化结果可复现 + # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 +QUANT_OBSERVER=minmax +QUANT_BATCHSIZE=1 +QUANT_STEP=32 +QUANT_SEED=42 +DISABLE_QUANT_LIST= +QUANT_EXIST_ONNX= diff --git a/models/cv/classification/convnext_base/ixrt/export.py b/models/cv/classification/convnext_base/ixrt/export.py new file mode 100644 index 0000000000000000000000000000000000000000..d9a2fe01e69085ca3c53782d32080c54cbbd1e5a --- /dev/null +++ b/models/cv/classification/convnext_base/ixrt/export.py @@ -0,0 +1,61 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import torch +import torchvision +import argparse + +def parse_args(): + parser = argparse.ArgumentParser() + + parser.add_argument("--weight", + type=str, + required=True, + help="pytorch model weight.") + + parser.add_argument("--output", + type=str, + required=True, + help="export onnx model path.") + + args = parser.parse_args() + return args + +def main(): + args = parse_args() + + model = torchvision.models.convnext_base() + model.load_state_dict(torch.load(args.weight)) + model.eval() + + input_names = ['input'] + output_names = ['output'] + dynamic_axes = {'input': {0: '-1'}, 'output': {0: '-1'}} + dummy_input = torch.randn(1, 3, 224, 224) + + torch.onnx.export( + model, + dummy_input, + args.output, + input_names = input_names, + dynamic_axes = dynamic_axes, + output_names = output_names, + opset_version=13 + ) + + print("Export onnx model successfully! ") + +if __name__ == "__main__": + main() diff --git a/models/cv/classification/convnext_base/ixrt/inference.py b/models/cv/classification/convnext_base/ixrt/inference.py new file mode 100644 index 0000000000000000000000000000000000000000..22f1644ced656c96602e15e468166d6df4fec92c --- /dev/null +++ b/models/cv/classification/convnext_base/ixrt/inference.py @@ -0,0 +1,171 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import argparse +import json +import os +import re +import time +from tqdm import tqdm + +import cv2 +import numpy as np +from cuda import cuda, cudart +import torch +import tensorrt + +from calibration_dataset import getdataloader +from common import eval_batch, create_engine_context, get_io_bindings + +def main(config): + dataloader = getdataloader(config.datasets_dir, config.loop_count, config.bsz, img_sz=config.imgsz) + + host_mem = tensorrt.IHostMemory + logger = tensorrt.Logger(tensorrt.Logger.ERROR) + + # Load Engine && I/O bindings + engine, context = create_engine_context(config.engine_file, logger) + inputs, outputs, allocations = get_io_bindings(engine) + + # Warm up + if config.warm_up > 0: + print("\nWarm Start.") + for i in range(config.warm_up): + context.execute_v2(allocations) + print("Warm Done.") + + # Inference + if config.test_mode == "FPS": + torch.cuda.synchronize() + start_time = time.time() + + for i in range(config.loop_count): + context.execute_v2(allocations) + + torch.cuda.synchronize() + end_time = time.time() + forward_time = end_time - start_time + + num_samples = 50000 + if config.loop_count * config.bsz < num_samples: + num_samples = config.loop_count * config.bsz + fps = num_samples / forward_time + + print("FPS : ", fps) + print(f"Performance Check : Test {fps} >= target {config.fps_target}") + if fps >= config.fps_target: + print("pass!") + exit() + else: + print("failed!") + exit(1) + + elif config.test_mode == "ACC": + + ## Prepare the output data + output = np.zeros(outputs[0]["shape"], outputs[0]["dtype"]) + print(f"output shape : {output.shape} output type : {output.dtype}") + + total_sample = 0 + acc_top1, acc_top5 = 0, 0 + + with tqdm(total= len(dataloader)) as _tqdm: + for idx, (batch_data, batch_label) in enumerate(dataloader): + batch_data = batch_data.numpy().astype(inputs[0]["dtype"]) + batch_data = np.ascontiguousarray(batch_data) + total_sample += batch_data.shape[0] + (err,) = cudart.cudaMemcpy( + inputs[0]["allocation"], + batch_data, + batch_data.nbytes, + cudart.cudaMemcpyKind.cudaMemcpyHostToDevice, + ) + assert err == cudart.cudaError_t.cudaSuccess + # cuda.memcpy_htod(inputs[0]["allocation"], batch_data) + context.execute_v2(allocations) + (err,) = cudart.cudaMemcpy( + output, + outputs[0]["allocation"], + outputs[0]["nbytes"], + cudart.cudaMemcpyKind.cudaMemcpyDeviceToHost, + ) + assert err == cudart.cudaError_t.cudaSuccess + # cuda.memcpy_dtoh(output, outputs[0]["allocation"]) + + # squeeze output shape [32,1000,1,1] to [32,1000] for mobilenet_v2 model + if len(output.shape) == 4: + output = output.squeeze(axis=(2,3)) + + batch_top1, batch_top5 = eval_batch(output, batch_label) + acc_top1 += batch_top1 + acc_top5 += batch_top5 + + _tqdm.set_postfix(acc_1='{:.4f}'.format(acc_top1/total_sample), + acc_5='{:.4f}'.format(acc_top5/total_sample)) + _tqdm.update(1) + err, = cudart.cudaFree(inputs[0]["allocation"]) + assert err == cudart.cudaError_t.cudaSuccess + err, = cudart.cudaFree(outputs[0]["allocation"]) + assert err == cudart.cudaError_t.cudaSuccess + print(F"Acc@1 : {acc_top1/total_sample} = {acc_top1}/{total_sample}") + print(F"Acc@5 : {acc_top5/total_sample} = {acc_top5}/{total_sample}") + acc1 = acc_top1/total_sample + print(f"Accuracy Check : Test {acc1} >= target {config.acc_target}") + if acc1 >= config.acc_target: + print("pass!") + exit() + else: + print("failed!") + exit(1) + +def parse_config(): + parser = argparse.ArgumentParser() + parser.add_argument("--test_mode", type=str, default="FPS", help="FPS MAP") + parser.add_argument( + "--engine_file", + type=str, + help="engine file path" + ) + parser.add_argument( + "--datasets_dir", + type=str, + default="", + help="ImageNet dir", + ) + parser.add_argument("--warm_up", type=int, default=-1, help="warm_up times") + parser.add_argument("--bsz", type=int, default=32, help="test batch size") + parser.add_argument( + "--imgsz", + "--img", + "--img-size", + type=int, + default=224, + help="inference size h,w", + ) + parser.add_argument("--use_async", action="store_true") + parser.add_argument( + "--device", type=int, default=0, help="cuda device, i.e. 0 or 0,1,2,3,4" + ) + parser.add_argument("--fps_target", type=float, default=-1.0) + parser.add_argument("--acc_target", type=float, default=-1.0) + parser.add_argument("--loop_count", type=int, default=-1) + + config = parser.parse_args() + return config + +if __name__ == "__main__": + config = parse_config() + main(config) diff --git a/models/cv/classification/convnext_base/ixrt/modify_batchsize.py b/models/cv/classification/convnext_base/ixrt/modify_batchsize.py new file mode 100644 index 0000000000000000000000000000000000000000..689b7a972dcbfec77c185592ede16bb4f04fa4fd --- /dev/null +++ b/models/cv/classification/convnext_base/ixrt/modify_batchsize.py @@ -0,0 +1,56 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import onnx +import argparse + +def change_input_dim(model, bsz): + batch_size = bsz + + # The following code changes the first dimension of every input to be batch_size + # Modify as appropriate ... note that this requires all inputs to + # have the same batch_size + inputs = model.graph.input + for input in inputs: + # Checks omitted.This assumes that all inputs are tensors and have a shape with first dim. + # Add checks as needed. + dim1 = input.type.tensor_type.shape.dim[0] + # update dim to be a symbolic value + if isinstance(batch_size, str): + # set dynamic batch size + dim1.dim_param = batch_size + elif (isinstance(batch_size, str) and batch_size.isdigit()) or isinstance(batch_size, int): + # set given batch size + dim1.dim_value = int(batch_size) + else: + # set batch size of 1 + dim1.dim_value = 1 + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--batch_size", type=int) + parser.add_argument("--origin_model", type=str) + parser.add_argument("--output_model", type=str) + args = parser.parse_args() + return args + +args = parse_args() +model = onnx.load(args.origin_model) +change_input_dim(model, args.batch_size) +onnx.save(model, args.output_model) + + + + + diff --git a/models/cv/classification/convnext_base/ixrt/scripts/infer_convnext_base_fp16_accuracy.sh b/models/cv/classification/convnext_base/ixrt/scripts/infer_convnext_base_fp16_accuracy.sh new file mode 100644 index 0000000000000000000000000000000000000000..a43c1a2067ef1527376279bc10ccace08dfa9b8f --- /dev/null +++ b/models/cv/classification/convnext_base/ixrt/scripts/infer_convnext_base_fp16_accuracy.sh @@ -0,0 +1,119 @@ +#!/bin/bash +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +EXIT_STATUS=0 +check_status() +{ + if ((${PIPESTATUS[0]} != 0));then + EXIT_STATUS=1 + fi +} + +# Run paraments +BSZ=32 +TGT=-1 +WARM_UP=0 +LOOP_COUNT=-1 +RUN_MODE=ACC +PRECISION=float16 + +# Update arguments +index=0 +options=$@ +arguments=($options) +for argument in $options +do + index=`expr $index + 1` + case $argument in + --bs) BSZ=${arguments[index]};; + --tgt) TGT=${arguments[index]};; + esac +done + +PROJ_DIR=${PROJ_DIR:-"."} +DATASETS_DIR="${DATASETS_DIR}" +CHECKPOINTS_DIR="${PROJ_DIR}" +RUN_DIR="${PROJ_DIR}" +CONFIG_DIR="${PROJ_DIR}/config/CONVNEXT_CONFIG" +source ${CONFIG_DIR} +ORIGINE_MODEL=${CHECKPOINTS_DIR}/${ORIGINE_MODEL} + +echo CHECKPOINTS_DIR : ${CHECKPOINTS_DIR} +echo DATASETS_DIR : ${DATASETS_DIR} +echo RUN_DIR : ${RUN_DIR} +echo CONFIG_DIR : ${CONFIG_DIR} +echo ====================== Model Info ====================== +echo Model Name : ${MODEL_NAME} +echo Onnx Path : ${ORIGINE_MODEL} + +step=0 +SIM_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_sim.onnx + +# Simplify Model +let step++ +echo; +echo [STEP ${step}] : Simplify Model +if [ -f ${SIM_MODEL} ];then + echo " "Simplify Model, ${SIM_MODEL} has been existed +else + python3 ${RUN_DIR}/simplify_model.py \ + --origin_model $ORIGINE_MODEL \ + --output_model ${SIM_MODEL} + echo " "Generate ${SIM_MODEL} +fi + +# Change Batchsize +let step++ +echo; +echo [STEP ${step}] : Change Batchsize +FINAL_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_${BSZ}.onnx +if [ -f $FINAL_MODEL ];then + echo " "Change Batchsize Skip, $FINAL_MODEL has been existed +else + python3 ${RUN_DIR}/modify_batchsize.py --batch_size ${BSZ} \ + --origin_model ${SIM_MODEL} --output_model ${FINAL_MODEL} + echo " "Generate ${FINAL_MODEL} +fi + +# Build Engine +let step++ +echo; +echo [STEP ${step}] : Build Engine +ENGINE_FILE=${CHECKPOINTS_DIR}/${MODEL_NAME}_${PRECISION}_bs${BSZ}.engine +if [ -f $ENGINE_FILE ];then + echo " "Build Engine Skip, $ENGINE_FILE has been existed +else + python3 ${RUN_DIR}/build_engine.py \ + --precision ${PRECISION} \ + --model ${FINAL_MODEL} \ + --engine ${ENGINE_FILE} + echo " "Generate Engine ${ENGINE_FILE} +fi + +# Inference +let step++ +echo; +echo [STEP ${step}] : Inference +python3 ${RUN_DIR}/inference.py \ + --engine_file=${ENGINE_FILE} \ + --datasets_dir=${DATASETS_DIR} \ + --imgsz=${IMGSIZE} \ + --warm_up=${WARM_UP} \ + --loop_count ${LOOP_COUNT} \ + --test_mode ${RUN_MODE} \ + --acc_target ${TGT} \ + --bsz ${BSZ}; check_status + +exit ${EXIT_STATUS} diff --git a/models/cv/classification/convnext_base/ixrt/scripts/infer_convnext_base_fp16_performance.sh b/models/cv/classification/convnext_base/ixrt/scripts/infer_convnext_base_fp16_performance.sh new file mode 100644 index 0000000000000000000000000000000000000000..3e5bca55207bccc0e09377816ba66e83d2909d86 --- /dev/null +++ b/models/cv/classification/convnext_base/ixrt/scripts/infer_convnext_base_fp16_performance.sh @@ -0,0 +1,119 @@ +#!/bin/bash +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +EXIT_STATUS=0 +check_status() +{ + if ((${PIPESTATUS[0]} != 0));then + EXIT_STATUS=1 + fi +} + +# Run paraments +BSZ=32 +TGT=-1 +WARM_UP=3 +LOOP_COUNT=20 +RUN_MODE=FPS +PRECISION=float16 + +# Update arguments +index=0 +options=$@ +arguments=($options) +for argument in $options +do + index=`expr $index + 1` + case $argument in + --bs) BSZ=${arguments[index]};; + --tgt) TGT=${arguments[index]};; + esac +done + +PROJ_DIR=${PROJ_DIR:-"."} +DATASETS_DIR="${DATASETS_DIR}" +CHECKPOINTS_DIR="${PROJ_DIR}" +RUN_DIR="${PROJ_DIR}" +CONFIG_DIR="${PROJ_DIR}/config/CONVNEXT_CONFIG" +source ${CONFIG_DIR} +ORIGINE_MODEL=${CHECKPOINTS_DIR}/${ORIGINE_MODEL} + +echo CHECKPOINTS_DIR : ${CHECKPOINTS_DIR} +echo DATASETS_DIR : ${DATASETS_DIR} +echo RUN_DIR : ${RUN_DIR} +echo CONFIG_DIR : ${CONFIG_DIR} +echo ====================== Model Info ====================== +echo Model Name : ${MODEL_NAME} +echo Onnx Path : ${ORIGINE_MODEL} + +step=0 +SIM_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_sim.onnx + +# Simplify Model +let step++ +echo; +echo [STEP ${step}] : Simplify Model +if [ -f ${SIM_MODEL} ];then + echo " "Simplify Model, ${SIM_MODEL} has been existed +else + python3 ${RUN_DIR}/simplify_model.py \ + --origin_model $ORIGINE_MODEL \ + --output_model ${SIM_MODEL} + echo " "Generate ${SIM_MODEL} +fi + +# Change Batchsize +let step++ +echo; +echo [STEP ${step}] : Change Batchsize +FINAL_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_${BSZ}.onnx +if [ -f $FINAL_MODEL ];then + echo " "Change Batchsize Skip, $FINAL_MODEL has been existed +else + python3 ${RUN_DIR}/modify_batchsize.py --batch_size ${BSZ} \ + --origin_model ${SIM_MODEL} --output_model ${FINAL_MODEL} + echo " "Generate ${FINAL_MODEL} +fi + +# Build Engine +let step++ +echo; +echo [STEP ${step}] : Build Engine +ENGINE_FILE=${CHECKPOINTS_DIR}/${MODEL_NAME}_${PRECISION}_bs${BSZ}.engine +if [ -f $ENGINE_FILE ];then + echo " "Build Engine Skip, $ENGINE_FILE has been existed +else + python3 ${RUN_DIR}/build_engine.py \ + --precision ${PRECISION} \ + --model ${FINAL_MODEL} \ + --engine ${ENGINE_FILE} + echo " "Generate Engine ${ENGINE_FILE} +fi + +# Inference +let step++ +echo; +echo [STEP ${step}] : Inference +python3 ${RUN_DIR}/inference.py \ + --engine_file=${ENGINE_FILE} \ + --datasets_dir=${DATASETS_DIR} \ + --imgsz=${IMGSIZE} \ + --warm_up=${WARM_UP} \ + --loop_count ${LOOP_COUNT} \ + --test_mode ${RUN_MODE} \ + --fps_target ${TGT} \ + --bsz ${BSZ}; check_status + +exit ${EXIT_STATUS} diff --git a/models/cv/classification/convnext_base/ixrt/simplify_model.py b/models/cv/classification/convnext_base/ixrt/simplify_model.py new file mode 100644 index 0000000000000000000000000000000000000000..9948a9fa083ff99ff88e556e96614b02cccaa965 --- /dev/null +++ b/models/cv/classification/convnext_base/ixrt/simplify_model.py @@ -0,0 +1,40 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import onnx +import argparse +from onnxsim import simplify + +# Simplify +def simplify_model(args): + onnx_model = onnx.load(args.origin_model) + model_simp, check = simplify(onnx_model) + model_simp = onnx.shape_inference.infer_shapes(model_simp) + onnx.save(model_simp, args.output_model) + print(" Simplify onnx Done.") + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--origin_model", type=str) + parser.add_argument("--output_model", type=str) + parser.add_argument("--reshape", action="store_true") + args = parser.parse_args() + return args + +args = parse_args() +simplify_model(args) + + + + diff --git a/models/cv/classification/deit_tiny/ixrt/README.md b/models/cv/classification/deit_tiny/ixrt/README.md new file mode 100644 index 0000000000000000000000000000000000000000..7a58584d058a99a036b9b201b0821cb03de7015b --- /dev/null +++ b/models/cv/classification/deit_tiny/ixrt/README.md @@ -0,0 +1,72 @@ +# DeiT-tiny + +## Description + +DeiT Tiny is a lightweight vision transformer designed for data-efficient learning. It achieves rapid training and high accuracy on small datasets through innovative attention distillation methods, while maintaining the simplicity and efficiency of the model. + +## Setup + +### Install + +```bash +# Install libGL +## CentOS +yum install -y mesa-libGL +## Ubuntu +apt install -y libgl1-mesa-dev + +pip3 install tqdm +pip3 install onnx +pip3 install onnxsim +pip3 install tabulate +pip3 install ppq +pip3 install tqdm +pip3 install cuda-python +``` + +### Download + +Pretrained model: + +Dataset: to download the validation dataset. + +### Model Conversion + +```bash +# git clone mmpretrain +git clone -b v0.24.0 https://github.com/open-mmlab/mmpretrain.git + +# export onnx model +python3 export.py --cfg mmpretrain/configs/deit/deit-tiny_pt-4xb256_in1k.py --weight deit-tiny_pt-4xb256_in1k_20220218-13b382a0.pth --output deit_tiny.onnx + +# Use onnxsim optimize onnx model +onnxsim deit_tiny.onnx deit_tiny_opt.onnx + +``` + +## Inference + +```bash +export DATASETS_DIR=/Path/to/imagenet_val/ +``` + +### FP16 + +```bash + +# Accuracy +bash scripts/infer_deit_tiny_fp16_accuracy.sh +# Performance +bash scripts/infer_deit_tiny_fp16_performance.sh + +``` + +## Results + +| Model | BatchSize | Precision | FPS | Top-1(%) | Top-5(%) | +| --------- | --------- | --------- | -------- | -------- | -------- | +| DeiT-tiny | 32 | FP16 | 1446.690 | 74.34 | 92.21 | + +## Reference + +Deit_tiny: \ No newline at end of file diff --git a/models/cv/classification/deit_tiny/ixrt/build_engine.py b/models/cv/classification/deit_tiny/ixrt/build_engine.py new file mode 100644 index 0000000000000000000000000000000000000000..038c15d5af7a51da6386aa0f994e830178330fb3 --- /dev/null +++ b/models/cv/classification/deit_tiny/ixrt/build_engine.py @@ -0,0 +1,52 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import cv2 +import argparse +import numpy as np + +import torch +import tensorrt + +def main(config): + IXRT_LOGGER = tensorrt.Logger(tensorrt.Logger.WARNING) + builder = tensorrt.Builder(IXRT_LOGGER) + EXPLICIT_BATCH = 1 << (int)(tensorrt.NetworkDefinitionCreationFlag.EXPLICIT_BATCH) + network = builder.create_network(EXPLICIT_BATCH) + build_config = builder.create_builder_config() + parser = tensorrt.OnnxParser(network, IXRT_LOGGER) + parser.parse_from_file(config.model) + + precision = tensorrt.BuilderFlag.INT8 if config.precision == "int8" else tensorrt.BuilderFlag.FP16 + build_config.set_flag(precision) + + plan = builder.build_serialized_network(network, build_config) + engine_file_path = config.engine + with open(engine_file_path, "wb") as f: + f.write(plan) + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--model", type=str) + parser.add_argument("--precision", type=str, choices=["float16", "int8", "float32"], default="int8", + help="The precision of datatype") + parser.add_argument("--engine", type=str, default=None) + args = parser.parse_args() + return args + +if __name__ == "__main__": + args = parse_args() + main(args) \ No newline at end of file diff --git a/models/cv/classification/deit_tiny/ixrt/calibration_dataset.py b/models/cv/classification/deit_tiny/ixrt/calibration_dataset.py new file mode 100644 index 0000000000000000000000000000000000000000..ec931c656abf5b2309dc9938490df46e4e8cdb19 --- /dev/null +++ b/models/cv/classification/deit_tiny/ixrt/calibration_dataset.py @@ -0,0 +1,112 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import os + +import torch +import torchvision.datasets +from torch.utils.data import DataLoader +from torchvision import models +from torchvision import transforms as T + + +class CalibrationImageNet(torchvision.datasets.ImageFolder): + def __init__(self, *args, **kwargs): + super(CalibrationImageNet, self).__init__(*args, **kwargs) + img2label_path = os.path.join(self.root, "val_map.txt") + if not os.path.exists(img2label_path): + raise FileNotFoundError(f"Not found label file `{img2label_path}`.") + + self.img2label_map = self.make_img2label_map(img2label_path) + + def make_img2label_map(self, path): + with open(path) as f: + lines = f.readlines() + + img2lable_map = dict() + for line in lines: + line = line.lstrip().rstrip().split("\t") + if len(line) != 2: + continue + img_name, label = line + img_name = img_name.strip() + if img_name in [None, ""]: + continue + label = int(label.strip()) + img2lable_map[img_name] = label + return img2lable_map + + def __getitem__(self, index): + path, target = self.samples[index] + sample = self.loader(path) + if self.transform is not None: + sample = self.transform(sample) + # if self.target_transform is not None: + # target = self.target_transform(target) + img_name = os.path.basename(path) + target = self.img2label_map[img_name] + + return sample, target + + +def create_dataloaders(data_path, num_samples=1024, img_sz=224, batch_size=2, workers=0): + dataset = CalibrationImageNet( + data_path, + transform=T.Compose( + [ + T.Resize(256), + T.CenterCrop(img_sz), + T.ToTensor(), + T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]), + ] + ), + ) + + calibration_dataset = dataset + if num_samples is not None: + calibration_dataset = torch.utils.data.Subset( + dataset, indices=range(num_samples) + ) + + calibration_dataloader = DataLoader( + calibration_dataset, + shuffle=True, + batch_size=batch_size, + drop_last=False, + num_workers=workers, + ) + + verify_dataloader = DataLoader( + dataset, + shuffle=False, + batch_size=batch_size, + drop_last=False, + num_workers=workers, + ) + + return calibration_dataloader, verify_dataloader + + +def getdataloader(dataset_dir, step=20, batch_size=32, workers=2, img_sz=224, total_sample=50000): + num_samples = min(total_sample, step * batch_size) + if step < 0: + num_samples = None + calibration_dataloader, _ = create_dataloaders( + dataset_dir, + img_sz=img_sz, + batch_size=batch_size, + workers=workers, + num_samples=num_samples, + ) + return calibration_dataloader diff --git a/models/cv/classification/deit_tiny/ixrt/common.py b/models/cv/classification/deit_tiny/ixrt/common.py new file mode 100644 index 0000000000000000000000000000000000000000..fd6a84d85fcc4ede8f64ed0e0a2dd7022527a956 --- /dev/null +++ b/models/cv/classification/deit_tiny/ixrt/common.py @@ -0,0 +1,81 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import cv2 +import glob +import torch +import tensorrt +import numpy as np +from cuda import cuda, cudart + +def eval_batch(batch_score, batch_label): + batch_score = torch.tensor(torch.from_numpy(batch_score), dtype=torch.float32) + values, indices = batch_score.topk(5) + top1, top5 = 0, 0 + for idx, label in enumerate(batch_label): + + if label == indices[idx][0]: + top1 += 1 + if label in indices[idx]: + top5 += 1 + return top1, top5 + +def create_engine_context(engine_path, logger): + with open(engine_path, "rb") as f: + runtime = tensorrt.Runtime(logger) + assert runtime + engine = runtime.deserialize_cuda_engine(f.read()) + assert engine + context = engine.create_execution_context() + assert context + + return engine, context + +def get_io_bindings(engine): + # Setup I/O bindings + inputs = [] + outputs = [] + allocations = [] + + for i in range(engine.num_bindings): + is_input = False + if engine.binding_is_input(i): + is_input = True + name = engine.get_binding_name(i) + dtype = engine.get_binding_dtype(i) + shape = engine.get_binding_shape(i) + if is_input: + batch_size = shape[0] + size = np.dtype(tensorrt.nptype(dtype)).itemsize + for s in shape: + size *= s + err, allocation = cudart.cudaMalloc(size) + assert err == cudart.cudaError_t.cudaSuccess + binding = { + "index": i, + "name": name, + "dtype": np.dtype(tensorrt.nptype(dtype)), + "shape": list(shape), + "allocation": allocation, + "nbytes": size, + } + print(f"binding {i}, name : {name} dtype : {np.dtype(tensorrt.nptype(dtype))} shape : {list(shape)}") + allocations.append(allocation) + if engine.binding_is_input(i): + inputs.append(binding) + else: + outputs.append(binding) + return inputs, outputs, allocations \ No newline at end of file diff --git a/models/cv/classification/deit_tiny/ixrt/config/DEIT_TINY_CONFIG b/models/cv/classification/deit_tiny/ixrt/config/DEIT_TINY_CONFIG new file mode 100644 index 0000000000000000000000000000000000000000..2464e4f5679a0a2beaab9f17d94390ee7769cf47 --- /dev/null +++ b/models/cv/classification/deit_tiny/ixrt/config/DEIT_TINY_CONFIG @@ -0,0 +1,33 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +# IMGSIZE : 模型输入hw大小 +# MODEL_NAME : 生成onnx/engine的basename +# ORIGINE_MODEL : 原始onnx文件名称 +IMGSIZE=224 +MODEL_NAME=deit_tiny +ORIGINE_MODEL=deit_tiny_opt.onnx + +# QUANT CONFIG (仅PRECISION为int8时生效) + # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] + # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) + # QUANT_STEP : 量化步数 + # QUANT_SEED : 随机种子 保证量化结果可复现 + # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 +QUANT_OBSERVER=minmax +QUANT_BATCHSIZE=1 +QUANT_STEP=32 +QUANT_SEED=42 +DISABLE_QUANT_LIST= +QUANT_EXIST_ONNX= diff --git a/models/cv/classification/deit_tiny/ixrt/export.py b/models/cv/classification/deit_tiny/ixrt/export.py new file mode 100644 index 0000000000000000000000000000000000000000..0078670ffbf4bbcce3358d4a2cedc42ce61176f5 --- /dev/null +++ b/models/cv/classification/deit_tiny/ixrt/export.py @@ -0,0 +1,78 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import argparse + +import torch +from mmcls.apis import init_model + +class Model(torch.nn.Module): + def __init__(self, config_file, checkpoint_file): + super().__init__() + self.model = init_model(config_file, checkpoint_file, device="cpu") + + def forward(self, x): + feat = self.model.backbone(x) + head = self.model.head.pre_logits(feat) + out_head = self.model.head.layers.head(head) + return out_head + +def parse_args(): + parser = argparse.ArgumentParser() + + parser.add_argument("--weight", + type=str, + required=True, + help="pytorch model weight.") + + parser.add_argument("--cfg", + type=str, + required=True, + help="model config file.") + + parser.add_argument("--output", + type=str, + required=True, + help="export onnx model path.") + + args = parser.parse_args() + return args + +def main(): + args = parse_args() + + config_file = args.cfg + checkpoint_file = args.weight + model = Model(config_file, checkpoint_file).eval() + + input_names = ['input'] + output_names = ['output'] + dynamic_axes = {'input': {0: '-1'}, 'output': {0: '-1'}} + dummy_input = torch.randn(1, 3, 224, 224) + + torch.onnx.export( + model, + dummy_input, + args.output, + input_names = input_names, + dynamic_axes = dynamic_axes, + output_names = output_names, + opset_version=13 + ) + + print("Export onnx model successfully! ") + +if __name__ == '__main__': + main() + diff --git a/models/cv/classification/deit_tiny/ixrt/inference.py b/models/cv/classification/deit_tiny/ixrt/inference.py new file mode 100644 index 0000000000000000000000000000000000000000..22f1644ced656c96602e15e468166d6df4fec92c --- /dev/null +++ b/models/cv/classification/deit_tiny/ixrt/inference.py @@ -0,0 +1,171 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import argparse +import json +import os +import re +import time +from tqdm import tqdm + +import cv2 +import numpy as np +from cuda import cuda, cudart +import torch +import tensorrt + +from calibration_dataset import getdataloader +from common import eval_batch, create_engine_context, get_io_bindings + +def main(config): + dataloader = getdataloader(config.datasets_dir, config.loop_count, config.bsz, img_sz=config.imgsz) + + host_mem = tensorrt.IHostMemory + logger = tensorrt.Logger(tensorrt.Logger.ERROR) + + # Load Engine && I/O bindings + engine, context = create_engine_context(config.engine_file, logger) + inputs, outputs, allocations = get_io_bindings(engine) + + # Warm up + if config.warm_up > 0: + print("\nWarm Start.") + for i in range(config.warm_up): + context.execute_v2(allocations) + print("Warm Done.") + + # Inference + if config.test_mode == "FPS": + torch.cuda.synchronize() + start_time = time.time() + + for i in range(config.loop_count): + context.execute_v2(allocations) + + torch.cuda.synchronize() + end_time = time.time() + forward_time = end_time - start_time + + num_samples = 50000 + if config.loop_count * config.bsz < num_samples: + num_samples = config.loop_count * config.bsz + fps = num_samples / forward_time + + print("FPS : ", fps) + print(f"Performance Check : Test {fps} >= target {config.fps_target}") + if fps >= config.fps_target: + print("pass!") + exit() + else: + print("failed!") + exit(1) + + elif config.test_mode == "ACC": + + ## Prepare the output data + output = np.zeros(outputs[0]["shape"], outputs[0]["dtype"]) + print(f"output shape : {output.shape} output type : {output.dtype}") + + total_sample = 0 + acc_top1, acc_top5 = 0, 0 + + with tqdm(total= len(dataloader)) as _tqdm: + for idx, (batch_data, batch_label) in enumerate(dataloader): + batch_data = batch_data.numpy().astype(inputs[0]["dtype"]) + batch_data = np.ascontiguousarray(batch_data) + total_sample += batch_data.shape[0] + (err,) = cudart.cudaMemcpy( + inputs[0]["allocation"], + batch_data, + batch_data.nbytes, + cudart.cudaMemcpyKind.cudaMemcpyHostToDevice, + ) + assert err == cudart.cudaError_t.cudaSuccess + # cuda.memcpy_htod(inputs[0]["allocation"], batch_data) + context.execute_v2(allocations) + (err,) = cudart.cudaMemcpy( + output, + outputs[0]["allocation"], + outputs[0]["nbytes"], + cudart.cudaMemcpyKind.cudaMemcpyDeviceToHost, + ) + assert err == cudart.cudaError_t.cudaSuccess + # cuda.memcpy_dtoh(output, outputs[0]["allocation"]) + + # squeeze output shape [32,1000,1,1] to [32,1000] for mobilenet_v2 model + if len(output.shape) == 4: + output = output.squeeze(axis=(2,3)) + + batch_top1, batch_top5 = eval_batch(output, batch_label) + acc_top1 += batch_top1 + acc_top5 += batch_top5 + + _tqdm.set_postfix(acc_1='{:.4f}'.format(acc_top1/total_sample), + acc_5='{:.4f}'.format(acc_top5/total_sample)) + _tqdm.update(1) + err, = cudart.cudaFree(inputs[0]["allocation"]) + assert err == cudart.cudaError_t.cudaSuccess + err, = cudart.cudaFree(outputs[0]["allocation"]) + assert err == cudart.cudaError_t.cudaSuccess + print(F"Acc@1 : {acc_top1/total_sample} = {acc_top1}/{total_sample}") + print(F"Acc@5 : {acc_top5/total_sample} = {acc_top5}/{total_sample}") + acc1 = acc_top1/total_sample + print(f"Accuracy Check : Test {acc1} >= target {config.acc_target}") + if acc1 >= config.acc_target: + print("pass!") + exit() + else: + print("failed!") + exit(1) + +def parse_config(): + parser = argparse.ArgumentParser() + parser.add_argument("--test_mode", type=str, default="FPS", help="FPS MAP") + parser.add_argument( + "--engine_file", + type=str, + help="engine file path" + ) + parser.add_argument( + "--datasets_dir", + type=str, + default="", + help="ImageNet dir", + ) + parser.add_argument("--warm_up", type=int, default=-1, help="warm_up times") + parser.add_argument("--bsz", type=int, default=32, help="test batch size") + parser.add_argument( + "--imgsz", + "--img", + "--img-size", + type=int, + default=224, + help="inference size h,w", + ) + parser.add_argument("--use_async", action="store_true") + parser.add_argument( + "--device", type=int, default=0, help="cuda device, i.e. 0 or 0,1,2,3,4" + ) + parser.add_argument("--fps_target", type=float, default=-1.0) + parser.add_argument("--acc_target", type=float, default=-1.0) + parser.add_argument("--loop_count", type=int, default=-1) + + config = parser.parse_args() + return config + +if __name__ == "__main__": + config = parse_config() + main(config) diff --git a/models/cv/classification/deit_tiny/ixrt/modify_batchsize.py b/models/cv/classification/deit_tiny/ixrt/modify_batchsize.py new file mode 100644 index 0000000000000000000000000000000000000000..689b7a972dcbfec77c185592ede16bb4f04fa4fd --- /dev/null +++ b/models/cv/classification/deit_tiny/ixrt/modify_batchsize.py @@ -0,0 +1,56 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import onnx +import argparse + +def change_input_dim(model, bsz): + batch_size = bsz + + # The following code changes the first dimension of every input to be batch_size + # Modify as appropriate ... note that this requires all inputs to + # have the same batch_size + inputs = model.graph.input + for input in inputs: + # Checks omitted.This assumes that all inputs are tensors and have a shape with first dim. + # Add checks as needed. + dim1 = input.type.tensor_type.shape.dim[0] + # update dim to be a symbolic value + if isinstance(batch_size, str): + # set dynamic batch size + dim1.dim_param = batch_size + elif (isinstance(batch_size, str) and batch_size.isdigit()) or isinstance(batch_size, int): + # set given batch size + dim1.dim_value = int(batch_size) + else: + # set batch size of 1 + dim1.dim_value = 1 + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--batch_size", type=int) + parser.add_argument("--origin_model", type=str) + parser.add_argument("--output_model", type=str) + args = parser.parse_args() + return args + +args = parse_args() +model = onnx.load(args.origin_model) +change_input_dim(model, args.batch_size) +onnx.save(model, args.output_model) + + + + + diff --git a/models/cv/classification/deit_tiny/ixrt/scripts/infer_deit_tiny_fp16_accuracy.sh b/models/cv/classification/deit_tiny/ixrt/scripts/infer_deit_tiny_fp16_accuracy.sh new file mode 100644 index 0000000000000000000000000000000000000000..0379dbe2f67a065abc0826cda83686aabb1cb34e --- /dev/null +++ b/models/cv/classification/deit_tiny/ixrt/scripts/infer_deit_tiny_fp16_accuracy.sh @@ -0,0 +1,119 @@ +#!/bin/bash +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +EXIT_STATUS=0 +check_status() +{ + if ((${PIPESTATUS[0]} != 0));then + EXIT_STATUS=1 + fi +} + +# Run paraments +BSZ=32 +TGT=-1 +WARM_UP=0 +LOOP_COUNT=-1 +RUN_MODE=ACC +PRECISION=float16 + +# Update arguments +index=0 +options=$@ +arguments=($options) +for argument in $options +do + index=`expr $index + 1` + case $argument in + --bs) BSZ=${arguments[index]};; + --tgt) TGT=${arguments[index]};; + esac +done + +PROJ_DIR=${PROJ_DIR:-"."} +DATASETS_DIR="${DATASETS_DIR}" +CHECKPOINTS_DIR="${PROJ_DIR}" +RUN_DIR="${PROJ_DIR}" +CONFIG_DIR="${PROJ_DIR}/config/DEIT_TINY_CONFIG" +source ${CONFIG_DIR} +ORIGINE_MODEL=${CHECKPOINTS_DIR}/${ORIGINE_MODEL} + +echo CHECKPOINTS_DIR : ${CHECKPOINTS_DIR} +echo DATASETS_DIR : ${DATASETS_DIR} +echo RUN_DIR : ${RUN_DIR} +echo CONFIG_DIR : ${CONFIG_DIR} +echo ====================== Model Info ====================== +echo Model Name : ${MODEL_NAME} +echo Onnx Path : ${ORIGINE_MODEL} + +step=0 +SIM_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_sim.onnx + +# Simplify Model +let step++ +echo; +echo [STEP ${step}] : Simplify Model +if [ -f ${SIM_MODEL} ];then + echo " "Simplify Model, ${SIM_MODEL} has been existed +else + python3 ${RUN_DIR}/simplify_model.py \ + --origin_model $ORIGINE_MODEL \ + --output_model ${SIM_MODEL} + echo " "Generate ${SIM_MODEL} +fi + +# Change Batchsize +let step++ +echo; +echo [STEP ${step}] : Change Batchsize +FINAL_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_${BSZ}.onnx +if [ -f $FINAL_MODEL ];then + echo " "Change Batchsize Skip, $FINAL_MODEL has been existed +else + python3 ${RUN_DIR}/modify_batchsize.py --batch_size ${BSZ} \ + --origin_model ${SIM_MODEL} --output_model ${FINAL_MODEL} + echo " "Generate ${FINAL_MODEL} +fi + +# Build Engine +let step++ +echo; +echo [STEP ${step}] : Build Engine +ENGINE_FILE=${CHECKPOINTS_DIR}/${MODEL_NAME}_${PRECISION}_bs${BSZ}.engine +if [ -f $ENGINE_FILE ];then + echo " "Build Engine Skip, $ENGINE_FILE has been existed +else + python3 ${RUN_DIR}/build_engine.py \ + --precision ${PRECISION} \ + --model ${FINAL_MODEL} \ + --engine ${ENGINE_FILE} + echo " "Generate Engine ${ENGINE_FILE} +fi + +# Inference +let step++ +echo; +echo [STEP ${step}] : Inference +python3 ${RUN_DIR}/inference.py \ + --engine_file=${ENGINE_FILE} \ + --datasets_dir=${DATASETS_DIR} \ + --imgsz=${IMGSIZE} \ + --warm_up=${WARM_UP} \ + --loop_count ${LOOP_COUNT} \ + --test_mode ${RUN_MODE} \ + --acc_target ${TGT} \ + --bsz ${BSZ}; check_status + +exit ${EXIT_STATUS} diff --git a/models/cv/classification/deit_tiny/ixrt/scripts/infer_deit_tiny_fp16_performance.sh b/models/cv/classification/deit_tiny/ixrt/scripts/infer_deit_tiny_fp16_performance.sh new file mode 100644 index 0000000000000000000000000000000000000000..18b48851daed8821def8500cf55b6387c32c3add --- /dev/null +++ b/models/cv/classification/deit_tiny/ixrt/scripts/infer_deit_tiny_fp16_performance.sh @@ -0,0 +1,119 @@ +#!/bin/bash +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +EXIT_STATUS=0 +check_status() +{ + if ((${PIPESTATUS[0]} != 0));then + EXIT_STATUS=1 + fi +} + +# Run paraments +BSZ=32 +TGT=-1 +WARM_UP=3 +LOOP_COUNT=20 +RUN_MODE=FPS +PRECISION=float16 + +# Update arguments +index=0 +options=$@ +arguments=($options) +for argument in $options +do + index=`expr $index + 1` + case $argument in + --bs) BSZ=${arguments[index]};; + --tgt) TGT=${arguments[index]};; + esac +done + +PROJ_DIR=${PROJ_DIR:-"."} +DATASETS_DIR="${DATASETS_DIR}" +CHECKPOINTS_DIR="${PROJ_DIR}" +RUN_DIR="${PROJ_DIR}" +CONFIG_DIR="${PROJ_DIR}/config/DEIT_TINY_CONFIG" +source ${CONFIG_DIR} +ORIGINE_MODEL=${CHECKPOINTS_DIR}/${ORIGINE_MODEL} + +echo CHECKPOINTS_DIR : ${CHECKPOINTS_DIR} +echo DATASETS_DIR : ${DATASETS_DIR} +echo RUN_DIR : ${RUN_DIR} +echo CONFIG_DIR : ${CONFIG_DIR} +echo ====================== Model Info ====================== +echo Model Name : ${MODEL_NAME} +echo Onnx Path : ${ORIGINE_MODEL} + +step=0 +SIM_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_sim.onnx + +# Simplify Model +let step++ +echo; +echo [STEP ${step}] : Simplify Model +if [ -f ${SIM_MODEL} ];then + echo " "Simplify Model, ${SIM_MODEL} has been existed +else + python3 ${RUN_DIR}/simplify_model.py \ + --origin_model $ORIGINE_MODEL \ + --output_model ${SIM_MODEL} + echo " "Generate ${SIM_MODEL} +fi + +# Change Batchsize +let step++ +echo; +echo [STEP ${step}] : Change Batchsize +FINAL_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_${BSZ}.onnx +if [ -f $FINAL_MODEL ];then + echo " "Change Batchsize Skip, $FINAL_MODEL has been existed +else + python3 ${RUN_DIR}/modify_batchsize.py --batch_size ${BSZ} \ + --origin_model ${SIM_MODEL} --output_model ${FINAL_MODEL} + echo " "Generate ${FINAL_MODEL} +fi + +# Build Engine +let step++ +echo; +echo [STEP ${step}] : Build Engine +ENGINE_FILE=${CHECKPOINTS_DIR}/${MODEL_NAME}_${PRECISION}_bs${BSZ}.engine +if [ -f $ENGINE_FILE ];then + echo " "Build Engine Skip, $ENGINE_FILE has been existed +else + python3 ${RUN_DIR}/build_engine.py \ + --precision ${PRECISION} \ + --model ${FINAL_MODEL} \ + --engine ${ENGINE_FILE} + echo " "Generate Engine ${ENGINE_FILE} +fi + +# Inference +let step++ +echo; +echo [STEP ${step}] : Inference +python3 ${RUN_DIR}/inference.py \ + --engine_file=${ENGINE_FILE} \ + --datasets_dir=${DATASETS_DIR} \ + --imgsz=${IMGSIZE} \ + --warm_up=${WARM_UP} \ + --loop_count ${LOOP_COUNT} \ + --test_mode ${RUN_MODE} \ + --fps_target ${TGT} \ + --bsz ${BSZ}; check_status + +exit ${EXIT_STATUS} diff --git a/models/cv/classification/deit_tiny/ixrt/simplify_model.py b/models/cv/classification/deit_tiny/ixrt/simplify_model.py new file mode 100644 index 0000000000000000000000000000000000000000..9948a9fa083ff99ff88e556e96614b02cccaa965 --- /dev/null +++ b/models/cv/classification/deit_tiny/ixrt/simplify_model.py @@ -0,0 +1,40 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import onnx +import argparse +from onnxsim import simplify + +# Simplify +def simplify_model(args): + onnx_model = onnx.load(args.origin_model) + model_simp, check = simplify(onnx_model) + model_simp = onnx.shape_inference.infer_shapes(model_simp) + onnx.save(model_simp, args.output_model) + print(" Simplify onnx Done.") + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--origin_model", type=str) + parser.add_argument("--output_model", type=str) + parser.add_argument("--reshape", action="store_true") + args = parser.parse_args() + return args + +args = parse_args() +simplify_model(args) + + + + diff --git a/models/cv/classification/densenet201/ixrt/README.md b/models/cv/classification/densenet201/ixrt/README.md new file mode 100644 index 0000000000000000000000000000000000000000..9475535534e4da7c0dc4710d51addbcaaec866d3 --- /dev/null +++ b/models/cv/classification/densenet201/ixrt/README.md @@ -0,0 +1,58 @@ +# DenseNet201 + +## Description + +DenseNet201 is a deep convolutional neural network that stands out for its unique dense connection architecture, where each layer integrates features from all previous layers, effectively reusing features and reducing the number of parameters. This design not only enhances the network's information flow and parameter efficiency but also increases the model's regularization effect, helping to prevent overfitting. DenseNet201 consists of multiple dense blocks and transition layers, capable of capturing rich feature representations while maintaining computational efficiency, making it suitable for complex image recognition tasks. + +## Setup + +### Install + +```bash +# Install libGL +## CentOS +yum install -y mesa-libGL +## Ubuntu +apt install -y libgl1-mesa-dev + +pip3 install tqdm +pip3 install onnx +pip3 install onnxsim +pip3 install tabulate +pip3 install ppq +pip3 install tqdm +pip3 install cuda-python +``` + +### Download + +Pretrained model: + +Dataset: to download the validation dataset. + +### Model Conversion + +```bash +python3 export.py --weight densenet201-c1103571.pth --output densenet201.onnx +``` + +## Inference + +```bash +export DATASETS_DIR=/Path/to/imagenet_val/ +``` + +### FP16 + +```bash +# Accuracy +bash scripts/infer_densenet201_fp16_accuracy.sh +# Performance +bash scripts/infer_densenet201_fp16_performance.sh +``` + +## Results + +| Model | BatchSize | Precision | FPS | Top-1(%) | Top-5(%) | +| ----------- | --------- | --------- | -------- | -------- | -------- | +| DenseNet201 | 32 | FP16 | 788.946 | 76.88 | 93.37 | diff --git a/models/cv/classification/densenet201/ixrt/build_engine.py b/models/cv/classification/densenet201/ixrt/build_engine.py new file mode 100644 index 0000000000000000000000000000000000000000..038c15d5af7a51da6386aa0f994e830178330fb3 --- /dev/null +++ b/models/cv/classification/densenet201/ixrt/build_engine.py @@ -0,0 +1,52 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import cv2 +import argparse +import numpy as np + +import torch +import tensorrt + +def main(config): + IXRT_LOGGER = tensorrt.Logger(tensorrt.Logger.WARNING) + builder = tensorrt.Builder(IXRT_LOGGER) + EXPLICIT_BATCH = 1 << (int)(tensorrt.NetworkDefinitionCreationFlag.EXPLICIT_BATCH) + network = builder.create_network(EXPLICIT_BATCH) + build_config = builder.create_builder_config() + parser = tensorrt.OnnxParser(network, IXRT_LOGGER) + parser.parse_from_file(config.model) + + precision = tensorrt.BuilderFlag.INT8 if config.precision == "int8" else tensorrt.BuilderFlag.FP16 + build_config.set_flag(precision) + + plan = builder.build_serialized_network(network, build_config) + engine_file_path = config.engine + with open(engine_file_path, "wb") as f: + f.write(plan) + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--model", type=str) + parser.add_argument("--precision", type=str, choices=["float16", "int8", "float32"], default="int8", + help="The precision of datatype") + parser.add_argument("--engine", type=str, default=None) + args = parser.parse_args() + return args + +if __name__ == "__main__": + args = parse_args() + main(args) \ No newline at end of file diff --git a/models/cv/classification/densenet201/ixrt/calibration_dataset.py b/models/cv/classification/densenet201/ixrt/calibration_dataset.py new file mode 100644 index 0000000000000000000000000000000000000000..ec931c656abf5b2309dc9938490df46e4e8cdb19 --- /dev/null +++ b/models/cv/classification/densenet201/ixrt/calibration_dataset.py @@ -0,0 +1,112 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import os + +import torch +import torchvision.datasets +from torch.utils.data import DataLoader +from torchvision import models +from torchvision import transforms as T + + +class CalibrationImageNet(torchvision.datasets.ImageFolder): + def __init__(self, *args, **kwargs): + super(CalibrationImageNet, self).__init__(*args, **kwargs) + img2label_path = os.path.join(self.root, "val_map.txt") + if not os.path.exists(img2label_path): + raise FileNotFoundError(f"Not found label file `{img2label_path}`.") + + self.img2label_map = self.make_img2label_map(img2label_path) + + def make_img2label_map(self, path): + with open(path) as f: + lines = f.readlines() + + img2lable_map = dict() + for line in lines: + line = line.lstrip().rstrip().split("\t") + if len(line) != 2: + continue + img_name, label = line + img_name = img_name.strip() + if img_name in [None, ""]: + continue + label = int(label.strip()) + img2lable_map[img_name] = label + return img2lable_map + + def __getitem__(self, index): + path, target = self.samples[index] + sample = self.loader(path) + if self.transform is not None: + sample = self.transform(sample) + # if self.target_transform is not None: + # target = self.target_transform(target) + img_name = os.path.basename(path) + target = self.img2label_map[img_name] + + return sample, target + + +def create_dataloaders(data_path, num_samples=1024, img_sz=224, batch_size=2, workers=0): + dataset = CalibrationImageNet( + data_path, + transform=T.Compose( + [ + T.Resize(256), + T.CenterCrop(img_sz), + T.ToTensor(), + T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]), + ] + ), + ) + + calibration_dataset = dataset + if num_samples is not None: + calibration_dataset = torch.utils.data.Subset( + dataset, indices=range(num_samples) + ) + + calibration_dataloader = DataLoader( + calibration_dataset, + shuffle=True, + batch_size=batch_size, + drop_last=False, + num_workers=workers, + ) + + verify_dataloader = DataLoader( + dataset, + shuffle=False, + batch_size=batch_size, + drop_last=False, + num_workers=workers, + ) + + return calibration_dataloader, verify_dataloader + + +def getdataloader(dataset_dir, step=20, batch_size=32, workers=2, img_sz=224, total_sample=50000): + num_samples = min(total_sample, step * batch_size) + if step < 0: + num_samples = None + calibration_dataloader, _ = create_dataloaders( + dataset_dir, + img_sz=img_sz, + batch_size=batch_size, + workers=workers, + num_samples=num_samples, + ) + return calibration_dataloader diff --git a/models/cv/classification/densenet201/ixrt/common.py b/models/cv/classification/densenet201/ixrt/common.py new file mode 100644 index 0000000000000000000000000000000000000000..fd6a84d85fcc4ede8f64ed0e0a2dd7022527a956 --- /dev/null +++ b/models/cv/classification/densenet201/ixrt/common.py @@ -0,0 +1,81 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import cv2 +import glob +import torch +import tensorrt +import numpy as np +from cuda import cuda, cudart + +def eval_batch(batch_score, batch_label): + batch_score = torch.tensor(torch.from_numpy(batch_score), dtype=torch.float32) + values, indices = batch_score.topk(5) + top1, top5 = 0, 0 + for idx, label in enumerate(batch_label): + + if label == indices[idx][0]: + top1 += 1 + if label in indices[idx]: + top5 += 1 + return top1, top5 + +def create_engine_context(engine_path, logger): + with open(engine_path, "rb") as f: + runtime = tensorrt.Runtime(logger) + assert runtime + engine = runtime.deserialize_cuda_engine(f.read()) + assert engine + context = engine.create_execution_context() + assert context + + return engine, context + +def get_io_bindings(engine): + # Setup I/O bindings + inputs = [] + outputs = [] + allocations = [] + + for i in range(engine.num_bindings): + is_input = False + if engine.binding_is_input(i): + is_input = True + name = engine.get_binding_name(i) + dtype = engine.get_binding_dtype(i) + shape = engine.get_binding_shape(i) + if is_input: + batch_size = shape[0] + size = np.dtype(tensorrt.nptype(dtype)).itemsize + for s in shape: + size *= s + err, allocation = cudart.cudaMalloc(size) + assert err == cudart.cudaError_t.cudaSuccess + binding = { + "index": i, + "name": name, + "dtype": np.dtype(tensorrt.nptype(dtype)), + "shape": list(shape), + "allocation": allocation, + "nbytes": size, + } + print(f"binding {i}, name : {name} dtype : {np.dtype(tensorrt.nptype(dtype))} shape : {list(shape)}") + allocations.append(allocation) + if engine.binding_is_input(i): + inputs.append(binding) + else: + outputs.append(binding) + return inputs, outputs, allocations \ No newline at end of file diff --git a/models/cv/classification/densenet201/ixrt/config/DENSENET201_CONFIG b/models/cv/classification/densenet201/ixrt/config/DENSENET201_CONFIG new file mode 100644 index 0000000000000000000000000000000000000000..df472521a3b2507c9f0547316b4a0f075736e885 --- /dev/null +++ b/models/cv/classification/densenet201/ixrt/config/DENSENET201_CONFIG @@ -0,0 +1,33 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +# IMGSIZE : 模型输入hw大小 +# MODEL_NAME : 生成onnx/engine的basename +# ORIGINE_MODEL : 原始onnx文件名称 +IMGSIZE=224 +MODEL_NAME=densenet201 +ORIGINE_MODEL=densenet201.onnx + +# QUANT CONFIG (仅PRECISION为int8时生效) + # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] + # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) + # QUANT_STEP : 量化步数 + # QUANT_SEED : 随机种子 保证量化结果可复现 + # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 +QUANT_OBSERVER=minmax +QUANT_BATCHSIZE=1 +QUANT_STEP=32 +QUANT_SEED=42 +DISABLE_QUANT_LIST= +QUANT_EXIST_ONNX= diff --git a/models/cv/classification/densenet201/ixrt/export.py b/models/cv/classification/densenet201/ixrt/export.py new file mode 100644 index 0000000000000000000000000000000000000000..66019547bf1101889f3fd699581741a114890959 --- /dev/null +++ b/models/cv/classification/densenet201/ixrt/export.py @@ -0,0 +1,74 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import torch +import torchvision +import argparse +import re + +def parse_args(): + parser = argparse.ArgumentParser() + + parser.add_argument("--weight", + type=str, + required=True, + help="pytorch model weight.") + + parser.add_argument("--output", + type=str, + required=True, + help="export onnx model path.") + + args = parser.parse_args() + return args + +def main(): + args = parse_args() + + model = torchvision.models.densenet201(weights=False) + + state_dict = torch.load(args.weight) + + pattern = re.compile(r'^(.*denselayer\d+\.(?:norm|relu|conv))\.((?:[12])\.(?:weight|bias|running_mean|running_var))$' + ) + for key in list(state_dict.keys()): + res = pattern.match(key) + if res: + new_key = res.group(1) + res.group(2) + state_dict[new_key] = state_dict[key] + del state_dict[key] + + model.load_state_dict(state_dict) + model.eval() + + input_names = ['input'] + output_names = ['output'] + dynamic_axes = {'input': {0: '-1'}, 'output': {0: '-1'}} + dummy_input = torch.randn(1, 3, 224, 224) + + torch.onnx.export( + model, + dummy_input, + args.output, + input_names = input_names, + dynamic_axes = dynamic_axes, + output_names = output_names, + opset_version=13 + ) + + print("Export onnx model successfully! ") + +if __name__ == "__main__": + main() diff --git a/models/cv/classification/densenet201/ixrt/inference.py b/models/cv/classification/densenet201/ixrt/inference.py new file mode 100644 index 0000000000000000000000000000000000000000..22f1644ced656c96602e15e468166d6df4fec92c --- /dev/null +++ b/models/cv/classification/densenet201/ixrt/inference.py @@ -0,0 +1,171 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import argparse +import json +import os +import re +import time +from tqdm import tqdm + +import cv2 +import numpy as np +from cuda import cuda, cudart +import torch +import tensorrt + +from calibration_dataset import getdataloader +from common import eval_batch, create_engine_context, get_io_bindings + +def main(config): + dataloader = getdataloader(config.datasets_dir, config.loop_count, config.bsz, img_sz=config.imgsz) + + host_mem = tensorrt.IHostMemory + logger = tensorrt.Logger(tensorrt.Logger.ERROR) + + # Load Engine && I/O bindings + engine, context = create_engine_context(config.engine_file, logger) + inputs, outputs, allocations = get_io_bindings(engine) + + # Warm up + if config.warm_up > 0: + print("\nWarm Start.") + for i in range(config.warm_up): + context.execute_v2(allocations) + print("Warm Done.") + + # Inference + if config.test_mode == "FPS": + torch.cuda.synchronize() + start_time = time.time() + + for i in range(config.loop_count): + context.execute_v2(allocations) + + torch.cuda.synchronize() + end_time = time.time() + forward_time = end_time - start_time + + num_samples = 50000 + if config.loop_count * config.bsz < num_samples: + num_samples = config.loop_count * config.bsz + fps = num_samples / forward_time + + print("FPS : ", fps) + print(f"Performance Check : Test {fps} >= target {config.fps_target}") + if fps >= config.fps_target: + print("pass!") + exit() + else: + print("failed!") + exit(1) + + elif config.test_mode == "ACC": + + ## Prepare the output data + output = np.zeros(outputs[0]["shape"], outputs[0]["dtype"]) + print(f"output shape : {output.shape} output type : {output.dtype}") + + total_sample = 0 + acc_top1, acc_top5 = 0, 0 + + with tqdm(total= len(dataloader)) as _tqdm: + for idx, (batch_data, batch_label) in enumerate(dataloader): + batch_data = batch_data.numpy().astype(inputs[0]["dtype"]) + batch_data = np.ascontiguousarray(batch_data) + total_sample += batch_data.shape[0] + (err,) = cudart.cudaMemcpy( + inputs[0]["allocation"], + batch_data, + batch_data.nbytes, + cudart.cudaMemcpyKind.cudaMemcpyHostToDevice, + ) + assert err == cudart.cudaError_t.cudaSuccess + # cuda.memcpy_htod(inputs[0]["allocation"], batch_data) + context.execute_v2(allocations) + (err,) = cudart.cudaMemcpy( + output, + outputs[0]["allocation"], + outputs[0]["nbytes"], + cudart.cudaMemcpyKind.cudaMemcpyDeviceToHost, + ) + assert err == cudart.cudaError_t.cudaSuccess + # cuda.memcpy_dtoh(output, outputs[0]["allocation"]) + + # squeeze output shape [32,1000,1,1] to [32,1000] for mobilenet_v2 model + if len(output.shape) == 4: + output = output.squeeze(axis=(2,3)) + + batch_top1, batch_top5 = eval_batch(output, batch_label) + acc_top1 += batch_top1 + acc_top5 += batch_top5 + + _tqdm.set_postfix(acc_1='{:.4f}'.format(acc_top1/total_sample), + acc_5='{:.4f}'.format(acc_top5/total_sample)) + _tqdm.update(1) + err, = cudart.cudaFree(inputs[0]["allocation"]) + assert err == cudart.cudaError_t.cudaSuccess + err, = cudart.cudaFree(outputs[0]["allocation"]) + assert err == cudart.cudaError_t.cudaSuccess + print(F"Acc@1 : {acc_top1/total_sample} = {acc_top1}/{total_sample}") + print(F"Acc@5 : {acc_top5/total_sample} = {acc_top5}/{total_sample}") + acc1 = acc_top1/total_sample + print(f"Accuracy Check : Test {acc1} >= target {config.acc_target}") + if acc1 >= config.acc_target: + print("pass!") + exit() + else: + print("failed!") + exit(1) + +def parse_config(): + parser = argparse.ArgumentParser() + parser.add_argument("--test_mode", type=str, default="FPS", help="FPS MAP") + parser.add_argument( + "--engine_file", + type=str, + help="engine file path" + ) + parser.add_argument( + "--datasets_dir", + type=str, + default="", + help="ImageNet dir", + ) + parser.add_argument("--warm_up", type=int, default=-1, help="warm_up times") + parser.add_argument("--bsz", type=int, default=32, help="test batch size") + parser.add_argument( + "--imgsz", + "--img", + "--img-size", + type=int, + default=224, + help="inference size h,w", + ) + parser.add_argument("--use_async", action="store_true") + parser.add_argument( + "--device", type=int, default=0, help="cuda device, i.e. 0 or 0,1,2,3,4" + ) + parser.add_argument("--fps_target", type=float, default=-1.0) + parser.add_argument("--acc_target", type=float, default=-1.0) + parser.add_argument("--loop_count", type=int, default=-1) + + config = parser.parse_args() + return config + +if __name__ == "__main__": + config = parse_config() + main(config) diff --git a/models/cv/classification/densenet201/ixrt/modify_batchsize.py b/models/cv/classification/densenet201/ixrt/modify_batchsize.py new file mode 100644 index 0000000000000000000000000000000000000000..689b7a972dcbfec77c185592ede16bb4f04fa4fd --- /dev/null +++ b/models/cv/classification/densenet201/ixrt/modify_batchsize.py @@ -0,0 +1,56 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import onnx +import argparse + +def change_input_dim(model, bsz): + batch_size = bsz + + # The following code changes the first dimension of every input to be batch_size + # Modify as appropriate ... note that this requires all inputs to + # have the same batch_size + inputs = model.graph.input + for input in inputs: + # Checks omitted.This assumes that all inputs are tensors and have a shape with first dim. + # Add checks as needed. + dim1 = input.type.tensor_type.shape.dim[0] + # update dim to be a symbolic value + if isinstance(batch_size, str): + # set dynamic batch size + dim1.dim_param = batch_size + elif (isinstance(batch_size, str) and batch_size.isdigit()) or isinstance(batch_size, int): + # set given batch size + dim1.dim_value = int(batch_size) + else: + # set batch size of 1 + dim1.dim_value = 1 + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--batch_size", type=int) + parser.add_argument("--origin_model", type=str) + parser.add_argument("--output_model", type=str) + args = parser.parse_args() + return args + +args = parse_args() +model = onnx.load(args.origin_model) +change_input_dim(model, args.batch_size) +onnx.save(model, args.output_model) + + + + + diff --git a/models/cv/classification/densenet201/ixrt/scripts/infer_densenet201_fp16_accuracy.sh b/models/cv/classification/densenet201/ixrt/scripts/infer_densenet201_fp16_accuracy.sh new file mode 100644 index 0000000000000000000000000000000000000000..30024cbc93039df2675d7cdfe25b373d8ca89de4 --- /dev/null +++ b/models/cv/classification/densenet201/ixrt/scripts/infer_densenet201_fp16_accuracy.sh @@ -0,0 +1,119 @@ +#!/bin/bash +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +EXIT_STATUS=0 +check_status() +{ + if ((${PIPESTATUS[0]} != 0));then + EXIT_STATUS=1 + fi +} + +# Run paraments +BSZ=32 +TGT=-1 +WARM_UP=0 +LOOP_COUNT=-1 +RUN_MODE=ACC +PRECISION=float16 + +# Update arguments +index=0 +options=$@ +arguments=($options) +for argument in $options +do + index=`expr $index + 1` + case $argument in + --bs) BSZ=${arguments[index]};; + --tgt) TGT=${arguments[index]};; + esac +done + +PROJ_DIR=${PROJ_DIR:-"."} +DATASETS_DIR="${DATASETS_DIR}" +CHECKPOINTS_DIR="${PROJ_DIR}" +RUN_DIR="${PROJ_DIR}" +CONFIG_DIR="${PROJ_DIR}/config/DENSENET201_CONFIG" +source ${CONFIG_DIR} +ORIGINE_MODEL=${CHECKPOINTS_DIR}/${ORIGINE_MODEL} + +echo CHECKPOINTS_DIR : ${CHECKPOINTS_DIR} +echo DATASETS_DIR : ${DATASETS_DIR} +echo RUN_DIR : ${RUN_DIR} +echo CONFIG_DIR : ${CONFIG_DIR} +echo ====================== Model Info ====================== +echo Model Name : ${MODEL_NAME} +echo Onnx Path : ${ORIGINE_MODEL} + +step=0 +SIM_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_sim.onnx + +# Simplify Model +let step++ +echo; +echo [STEP ${step}] : Simplify Model +if [ -f ${SIM_MODEL} ];then + echo " "Simplify Model, ${SIM_MODEL} has been existed +else + python3 ${RUN_DIR}/simplify_model.py \ + --origin_model $ORIGINE_MODEL \ + --output_model ${SIM_MODEL} + echo " "Generate ${SIM_MODEL} +fi + +# Change Batchsize +let step++ +echo; +echo [STEP ${step}] : Change Batchsize +FINAL_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_${BSZ}.onnx +if [ -f $FINAL_MODEL ];then + echo " "Change Batchsize Skip, $FINAL_MODEL has been existed +else + python3 ${RUN_DIR}/modify_batchsize.py --batch_size ${BSZ} \ + --origin_model ${SIM_MODEL} --output_model ${FINAL_MODEL} + echo " "Generate ${FINAL_MODEL} +fi + +# Build Engine +let step++ +echo; +echo [STEP ${step}] : Build Engine +ENGINE_FILE=${CHECKPOINTS_DIR}/${MODEL_NAME}_${PRECISION}_bs${BSZ}.engine +if [ -f $ENGINE_FILE ];then + echo " "Build Engine Skip, $ENGINE_FILE has been existed +else + python3 ${RUN_DIR}/build_engine.py \ + --precision ${PRECISION} \ + --model ${FINAL_MODEL} \ + --engine ${ENGINE_FILE} + echo " "Generate Engine ${ENGINE_FILE} +fi + +# Inference +let step++ +echo; +echo [STEP ${step}] : Inference +python3 ${RUN_DIR}/inference.py \ + --engine_file=${ENGINE_FILE} \ + --datasets_dir=${DATASETS_DIR} \ + --imgsz=${IMGSIZE} \ + --warm_up=${WARM_UP} \ + --loop_count ${LOOP_COUNT} \ + --test_mode ${RUN_MODE} \ + --acc_target ${TGT} \ + --bsz ${BSZ}; check_status + +exit ${EXIT_STATUS} diff --git a/models/cv/classification/densenet201/ixrt/scripts/infer_densenet201_fp16_performance.sh b/models/cv/classification/densenet201/ixrt/scripts/infer_densenet201_fp16_performance.sh new file mode 100644 index 0000000000000000000000000000000000000000..42b81e12d6c9e0b3be289ae1562cd39538335fd2 --- /dev/null +++ b/models/cv/classification/densenet201/ixrt/scripts/infer_densenet201_fp16_performance.sh @@ -0,0 +1,119 @@ +#!/bin/bash +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +EXIT_STATUS=0 +check_status() +{ + if ((${PIPESTATUS[0]} != 0));then + EXIT_STATUS=1 + fi +} + +# Run paraments +BSZ=32 +TGT=-1 +WARM_UP=3 +LOOP_COUNT=20 +RUN_MODE=FPS +PRECISION=float16 + +# Update arguments +index=0 +options=$@ +arguments=($options) +for argument in $options +do + index=`expr $index + 1` + case $argument in + --bs) BSZ=${arguments[index]};; + --tgt) TGT=${arguments[index]};; + esac +done + +PROJ_DIR=${PROJ_DIR:-"."} +DATASETS_DIR="${DATASETS_DIR}" +CHECKPOINTS_DIR="${PROJ_DIR}" +RUN_DIR="${PROJ_DIR}" +CONFIG_DIR="${PROJ_DIR}/config/DENSENET201_CONFIG" +source ${CONFIG_DIR} +ORIGINE_MODEL=${CHECKPOINTS_DIR}/${ORIGINE_MODEL} + +echo CHECKPOINTS_DIR : ${CHECKPOINTS_DIR} +echo DATASETS_DIR : ${DATASETS_DIR} +echo RUN_DIR : ${RUN_DIR} +echo CONFIG_DIR : ${CONFIG_DIR} +echo ====================== Model Info ====================== +echo Model Name : ${MODEL_NAME} +echo Onnx Path : ${ORIGINE_MODEL} + +step=0 +SIM_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_sim.onnx + +# Simplify Model +let step++ +echo; +echo [STEP ${step}] : Simplify Model +if [ -f ${SIM_MODEL} ];then + echo " "Simplify Model, ${SIM_MODEL} has been existed +else + python3 ${RUN_DIR}/simplify_model.py \ + --origin_model $ORIGINE_MODEL \ + --output_model ${SIM_MODEL} + echo " "Generate ${SIM_MODEL} +fi + +# Change Batchsize +let step++ +echo; +echo [STEP ${step}] : Change Batchsize +FINAL_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_${BSZ}.onnx +if [ -f $FINAL_MODEL ];then + echo " "Change Batchsize Skip, $FINAL_MODEL has been existed +else + python3 ${RUN_DIR}/modify_batchsize.py --batch_size ${BSZ} \ + --origin_model ${SIM_MODEL} --output_model ${FINAL_MODEL} + echo " "Generate ${FINAL_MODEL} +fi + +# Build Engine +let step++ +echo; +echo [STEP ${step}] : Build Engine +ENGINE_FILE=${CHECKPOINTS_DIR}/${MODEL_NAME}_${PRECISION}_bs${BSZ}.engine +if [ -f $ENGINE_FILE ];then + echo " "Build Engine Skip, $ENGINE_FILE has been existed +else + python3 ${RUN_DIR}/build_engine.py \ + --precision ${PRECISION} \ + --model ${FINAL_MODEL} \ + --engine ${ENGINE_FILE} + echo " "Generate Engine ${ENGINE_FILE} +fi + +# Inference +let step++ +echo; +echo [STEP ${step}] : Inference +python3 ${RUN_DIR}/inference.py \ + --engine_file=${ENGINE_FILE} \ + --datasets_dir=${DATASETS_DIR} \ + --imgsz=${IMGSIZE} \ + --warm_up=${WARM_UP} \ + --loop_count ${LOOP_COUNT} \ + --test_mode ${RUN_MODE} \ + --fps_target ${TGT} \ + --bsz ${BSZ}; check_status + +exit ${EXIT_STATUS} diff --git a/models/cv/classification/densenet201/ixrt/simplify_model.py b/models/cv/classification/densenet201/ixrt/simplify_model.py new file mode 100644 index 0000000000000000000000000000000000000000..9948a9fa083ff99ff88e556e96614b02cccaa965 --- /dev/null +++ b/models/cv/classification/densenet201/ixrt/simplify_model.py @@ -0,0 +1,40 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import onnx +import argparse +from onnxsim import simplify + +# Simplify +def simplify_model(args): + onnx_model = onnx.load(args.origin_model) + model_simp, check = simplify(onnx_model) + model_simp = onnx.shape_inference.infer_shapes(model_simp) + onnx.save(model_simp, args.output_model) + print(" Simplify onnx Done.") + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--origin_model", type=str) + parser.add_argument("--output_model", type=str) + parser.add_argument("--reshape", action="store_true") + args = parser.parse_args() + return args + +args = parse_args() +simplify_model(args) + + + + diff --git a/models/cv/classification/efficientnetv2_rw_t/ixrt/README.md b/models/cv/classification/efficientnetv2_rw_t/ixrt/README.md new file mode 100644 index 0000000000000000000000000000000000000000..bc6836d029c2b1b8312e5fd363d1075158dfd910 --- /dev/null +++ b/models/cv/classification/efficientnetv2_rw_t/ixrt/README.md @@ -0,0 +1,59 @@ +# EfficientNetv2_rw_t + +## Description + +EfficientNetV2_rw_t is an enhanced version of the EfficientNet family of convolutional neural network architectures. It builds upon the success of its predecessors by introducing novel advancements aimed at further improving performance and efficiency in various computer vision tasks. + +## Setup + +### Install + +```bash +# Install libGL +## CentOS +yum install -y mesa-libGL +## Ubuntu +apt install -y libgl1-mesa-dev + +pip3 install tqdm +pip3 install timm +pip3 install onnx +pip3 install onnxsim +pip3 install tabulate +pip3 install ppq +pip3 install tqdm +pip3 install cuda-python +``` + +### Download + +Pretrained model: + +Dataset: to download the validation dataset. + +### Model Conversion + +```bash +python3 export.py --weight efficientnetv2_t_agc-3620981a.pth --output efficientnetv2_rw_t.onnx +``` + +## Inference + +```bash +export DATASETS_DIR=/Path/to/imagenet_val/ +``` + +### FP16 + +```bash +# Accuracy +bash scripts/infer_efficientnetv2_rw_t_fp16_accuracy.sh +# Performance +bash scripts/infer_efficientnetv2_rw_t_fp16_performance.sh +``` + +## Results + +Model |BatchSize |Precision |FPS |Top-1(%) |Top-5(%) +--------------------|-----------|----------|---------|---------|-------- +Efficientnetv2_rw_t | 32 | FP16 | 1525.22 | 82.336 | 96.194 diff --git a/models/cv/classification/efficientnetv2_rw_t/ixrt/build_engine.py b/models/cv/classification/efficientnetv2_rw_t/ixrt/build_engine.py new file mode 100644 index 0000000000000000000000000000000000000000..038c15d5af7a51da6386aa0f994e830178330fb3 --- /dev/null +++ b/models/cv/classification/efficientnetv2_rw_t/ixrt/build_engine.py @@ -0,0 +1,52 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import cv2 +import argparse +import numpy as np + +import torch +import tensorrt + +def main(config): + IXRT_LOGGER = tensorrt.Logger(tensorrt.Logger.WARNING) + builder = tensorrt.Builder(IXRT_LOGGER) + EXPLICIT_BATCH = 1 << (int)(tensorrt.NetworkDefinitionCreationFlag.EXPLICIT_BATCH) + network = builder.create_network(EXPLICIT_BATCH) + build_config = builder.create_builder_config() + parser = tensorrt.OnnxParser(network, IXRT_LOGGER) + parser.parse_from_file(config.model) + + precision = tensorrt.BuilderFlag.INT8 if config.precision == "int8" else tensorrt.BuilderFlag.FP16 + build_config.set_flag(precision) + + plan = builder.build_serialized_network(network, build_config) + engine_file_path = config.engine + with open(engine_file_path, "wb") as f: + f.write(plan) + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--model", type=str) + parser.add_argument("--precision", type=str, choices=["float16", "int8", "float32"], default="int8", + help="The precision of datatype") + parser.add_argument("--engine", type=str, default=None) + args = parser.parse_args() + return args + +if __name__ == "__main__": + args = parse_args() + main(args) \ No newline at end of file diff --git a/models/cv/classification/efficientnetv2_rw_t/ixrt/calibration_dataset.py b/models/cv/classification/efficientnetv2_rw_t/ixrt/calibration_dataset.py new file mode 100644 index 0000000000000000000000000000000000000000..ec931c656abf5b2309dc9938490df46e4e8cdb19 --- /dev/null +++ b/models/cv/classification/efficientnetv2_rw_t/ixrt/calibration_dataset.py @@ -0,0 +1,112 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import os + +import torch +import torchvision.datasets +from torch.utils.data import DataLoader +from torchvision import models +from torchvision import transforms as T + + +class CalibrationImageNet(torchvision.datasets.ImageFolder): + def __init__(self, *args, **kwargs): + super(CalibrationImageNet, self).__init__(*args, **kwargs) + img2label_path = os.path.join(self.root, "val_map.txt") + if not os.path.exists(img2label_path): + raise FileNotFoundError(f"Not found label file `{img2label_path}`.") + + self.img2label_map = self.make_img2label_map(img2label_path) + + def make_img2label_map(self, path): + with open(path) as f: + lines = f.readlines() + + img2lable_map = dict() + for line in lines: + line = line.lstrip().rstrip().split("\t") + if len(line) != 2: + continue + img_name, label = line + img_name = img_name.strip() + if img_name in [None, ""]: + continue + label = int(label.strip()) + img2lable_map[img_name] = label + return img2lable_map + + def __getitem__(self, index): + path, target = self.samples[index] + sample = self.loader(path) + if self.transform is not None: + sample = self.transform(sample) + # if self.target_transform is not None: + # target = self.target_transform(target) + img_name = os.path.basename(path) + target = self.img2label_map[img_name] + + return sample, target + + +def create_dataloaders(data_path, num_samples=1024, img_sz=224, batch_size=2, workers=0): + dataset = CalibrationImageNet( + data_path, + transform=T.Compose( + [ + T.Resize(256), + T.CenterCrop(img_sz), + T.ToTensor(), + T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]), + ] + ), + ) + + calibration_dataset = dataset + if num_samples is not None: + calibration_dataset = torch.utils.data.Subset( + dataset, indices=range(num_samples) + ) + + calibration_dataloader = DataLoader( + calibration_dataset, + shuffle=True, + batch_size=batch_size, + drop_last=False, + num_workers=workers, + ) + + verify_dataloader = DataLoader( + dataset, + shuffle=False, + batch_size=batch_size, + drop_last=False, + num_workers=workers, + ) + + return calibration_dataloader, verify_dataloader + + +def getdataloader(dataset_dir, step=20, batch_size=32, workers=2, img_sz=224, total_sample=50000): + num_samples = min(total_sample, step * batch_size) + if step < 0: + num_samples = None + calibration_dataloader, _ = create_dataloaders( + dataset_dir, + img_sz=img_sz, + batch_size=batch_size, + workers=workers, + num_samples=num_samples, + ) + return calibration_dataloader diff --git a/models/cv/classification/efficientnetv2_rw_t/ixrt/common.py b/models/cv/classification/efficientnetv2_rw_t/ixrt/common.py new file mode 100644 index 0000000000000000000000000000000000000000..fd6a84d85fcc4ede8f64ed0e0a2dd7022527a956 --- /dev/null +++ b/models/cv/classification/efficientnetv2_rw_t/ixrt/common.py @@ -0,0 +1,81 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import cv2 +import glob +import torch +import tensorrt +import numpy as np +from cuda import cuda, cudart + +def eval_batch(batch_score, batch_label): + batch_score = torch.tensor(torch.from_numpy(batch_score), dtype=torch.float32) + values, indices = batch_score.topk(5) + top1, top5 = 0, 0 + for idx, label in enumerate(batch_label): + + if label == indices[idx][0]: + top1 += 1 + if label in indices[idx]: + top5 += 1 + return top1, top5 + +def create_engine_context(engine_path, logger): + with open(engine_path, "rb") as f: + runtime = tensorrt.Runtime(logger) + assert runtime + engine = runtime.deserialize_cuda_engine(f.read()) + assert engine + context = engine.create_execution_context() + assert context + + return engine, context + +def get_io_bindings(engine): + # Setup I/O bindings + inputs = [] + outputs = [] + allocations = [] + + for i in range(engine.num_bindings): + is_input = False + if engine.binding_is_input(i): + is_input = True + name = engine.get_binding_name(i) + dtype = engine.get_binding_dtype(i) + shape = engine.get_binding_shape(i) + if is_input: + batch_size = shape[0] + size = np.dtype(tensorrt.nptype(dtype)).itemsize + for s in shape: + size *= s + err, allocation = cudart.cudaMalloc(size) + assert err == cudart.cudaError_t.cudaSuccess + binding = { + "index": i, + "name": name, + "dtype": np.dtype(tensorrt.nptype(dtype)), + "shape": list(shape), + "allocation": allocation, + "nbytes": size, + } + print(f"binding {i}, name : {name} dtype : {np.dtype(tensorrt.nptype(dtype))} shape : {list(shape)}") + allocations.append(allocation) + if engine.binding_is_input(i): + inputs.append(binding) + else: + outputs.append(binding) + return inputs, outputs, allocations \ No newline at end of file diff --git a/models/cv/classification/efficientnetv2_rw_t/ixrt/config/EFFICIENTNETV2_CONFIG b/models/cv/classification/efficientnetv2_rw_t/ixrt/config/EFFICIENTNETV2_CONFIG new file mode 100644 index 0000000000000000000000000000000000000000..f3343fd3eab2d2d3fbc660ec1a94e4932a69c90f --- /dev/null +++ b/models/cv/classification/efficientnetv2_rw_t/ixrt/config/EFFICIENTNETV2_CONFIG @@ -0,0 +1,33 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +# IMGSIZE : 模型输入hw大小 +# MODEL_NAME : 生成onnx/engine的basename +# ORIGINE_MODEL : 原始onnx文件名称 +IMGSIZE=288 +MODEL_NAME=efficientnetv2_rw_t +ORIGINE_MODEL=efficientnetv2_rw_t.onnx + +# QUANT CONFIG (仅PRECISION为int8时生效) + # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] + # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) + # QUANT_STEP : 量化步数 + # QUANT_SEED : 随机种子 保证量化结果可复现 + # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 +QUANT_OBSERVER=minmax +QUANT_BATCHSIZE=1 +QUANT_STEP=32 +QUANT_SEED=42 +DISABLE_QUANT_LIST= +QUANT_EXIST_ONNX= diff --git a/models/cv/classification/efficientnetv2_rw_t/ixrt/export.py b/models/cv/classification/efficientnetv2_rw_t/ixrt/export.py new file mode 100644 index 0000000000000000000000000000000000000000..3c3f579c02f06cec8be82f7ca1a797ecbab960c3 --- /dev/null +++ b/models/cv/classification/efficientnetv2_rw_t/ixrt/export.py @@ -0,0 +1,58 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import timm +import torch +import argparse + +def parse_args(): + parser = argparse.ArgumentParser() + + parser.add_argument("--weight", + type=str, + required=True, + help="pytorch model weight.") + + parser.add_argument("--output", + type=str, + required=True, + help="export onnx model path.") + + args = parser.parse_args() + return args + +def main(): + args = parse_args() + + model = timm.create_model('efficientnetv2_rw_t', checkpoint_path=args.weight) + model.eval() + + dummy_input = torch.randn([32, 3, 288, 288]) + + torch.onnx.export( + model, + dummy_input, + args.output, + opset_version=13, + do_constant_folding=True, + input_names=["input"], + output_names=["output"], + dynamic_axes={"input": {0: "batch"}, "output": {0: "batch"}} + ) + + print("Export onnx model successfully! ") + +if __name__ == "__main__": + main() \ No newline at end of file diff --git a/models/cv/classification/efficientnetv2_rw_t/ixrt/inference.py b/models/cv/classification/efficientnetv2_rw_t/ixrt/inference.py new file mode 100644 index 0000000000000000000000000000000000000000..e04f1c3aead1aae2aec278cce6a60d49b9687ee0 --- /dev/null +++ b/models/cv/classification/efficientnetv2_rw_t/ixrt/inference.py @@ -0,0 +1,187 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import argparse +import json +import os +import re +import time +from tqdm import tqdm + +import cv2 +import numpy as np +from cuda import cuda, cudart +import torch +import tensorrt + +from timm.data import create_dataset, create_loader +from common import eval_batch, create_engine_context, get_io_bindings + +def get_dataloader(data_path, batch_size, num_workers): + datasets = create_dataset(root=data_path, name="") + + dataloader = create_loader( + datasets, + input_size=(3, 288, 288), + batch_size=batch_size, + interpolation='bicubic', + mean=(0.485, 0.456, 0.406), + std=(0.229, 0.224, 0.225), + crop_pct=1.0, + use_prefetcher = False, + num_workers = num_workers + ) + return dataloader + +def main(config): + dataloader = get_dataloader(config.datasets_dir, config.bsz, 16) + + host_mem = tensorrt.IHostMemory + logger = tensorrt.Logger(tensorrt.Logger.ERROR) + + # Load Engine && I/O bindings + engine, context = create_engine_context(config.engine_file, logger) + inputs, outputs, allocations = get_io_bindings(engine) + + # Warm up + if config.warm_up > 0: + print("\nWarm Start.") + for i in range(config.warm_up): + context.execute_v2(allocations) + print("Warm Done.") + + # Inference + if config.test_mode == "FPS": + torch.cuda.synchronize() + start_time = time.time() + + for i in range(config.loop_count): + context.execute_v2(allocations) + + torch.cuda.synchronize() + end_time = time.time() + forward_time = end_time - start_time + + num_samples = 50000 + if config.loop_count * config.bsz < num_samples: + num_samples = config.loop_count * config.bsz + fps = num_samples / forward_time + + print("FPS : ", fps) + print(f"Performance Check : Test {fps} >= target {config.fps_target}") + if fps >= config.fps_target: + print("pass!") + exit() + else: + print("failed!") + exit(1) + + elif config.test_mode == "ACC": + + ## Prepare the output data + output = np.zeros(outputs[0]["shape"], outputs[0]["dtype"]) + print(f"output shape : {output.shape} output type : {output.dtype}") + + total_sample = 0 + acc_top1, acc_top5 = 0, 0 + + with tqdm(total= len(dataloader)) as _tqdm: + for idx, (batch_data, batch_label) in enumerate(dataloader): + batch_data = batch_data.numpy().astype(inputs[0]["dtype"]) + batch_data = np.ascontiguousarray(batch_data) + total_sample += batch_data.shape[0] + (err,) = cudart.cudaMemcpy( + inputs[0]["allocation"], + batch_data, + batch_data.nbytes, + cudart.cudaMemcpyKind.cudaMemcpyHostToDevice, + ) + assert err == cudart.cudaError_t.cudaSuccess + # cuda.memcpy_htod(inputs[0]["allocation"], batch_data) + context.execute_v2(allocations) + (err,) = cudart.cudaMemcpy( + output, + outputs[0]["allocation"], + outputs[0]["nbytes"], + cudart.cudaMemcpyKind.cudaMemcpyDeviceToHost, + ) + assert err == cudart.cudaError_t.cudaSuccess + # cuda.memcpy_dtoh(output, outputs[0]["allocation"]) + + # squeeze output shape [32,1000,1,1] to [32,1000] for mobilenet_v2 model + if len(output.shape) == 4: + output = output.squeeze(axis=(2,3)) + + batch_top1, batch_top5 = eval_batch(output, batch_label) + acc_top1 += batch_top1 + acc_top5 += batch_top5 + + _tqdm.set_postfix(acc_1='{:.4f}'.format(acc_top1/total_sample), + acc_5='{:.4f}'.format(acc_top5/total_sample)) + _tqdm.update(1) + err, = cudart.cudaFree(inputs[0]["allocation"]) + assert err == cudart.cudaError_t.cudaSuccess + err, = cudart.cudaFree(outputs[0]["allocation"]) + assert err == cudart.cudaError_t.cudaSuccess + print(F"Acc@1 : {acc_top1/total_sample} = {acc_top1}/{total_sample}") + print(F"Acc@5 : {acc_top5/total_sample} = {acc_top5}/{total_sample}") + acc1 = acc_top1/total_sample + print(f"Accuracy Check : Test {acc1} >= target {config.acc_target}") + if acc1 >= config.acc_target: + print("pass!") + exit() + else: + print("failed!") + exit(1) + +def parse_config(): + parser = argparse.ArgumentParser() + parser.add_argument("--test_mode", type=str, default="FPS", help="FPS MAP") + parser.add_argument( + "--engine_file", + type=str, + help="engine file path" + ) + parser.add_argument( + "--datasets_dir", + type=str, + default="", + help="ImageNet dir", + ) + parser.add_argument("--warm_up", type=int, default=-1, help="warm_up times") + parser.add_argument("--bsz", type=int, default=32, help="test batch size") + parser.add_argument( + "--imgsz", + "--img", + "--img-size", + type=int, + default=224, + help="inference size h,w", + ) + parser.add_argument("--use_async", action="store_true") + parser.add_argument( + "--device", type=int, default=0, help="cuda device, i.e. 0 or 0,1,2,3,4" + ) + parser.add_argument("--fps_target", type=float, default=-1.0) + parser.add_argument("--acc_target", type=float, default=-1.0) + parser.add_argument("--loop_count", type=int, default=-1) + + config = parser.parse_args() + return config + +if __name__ == "__main__": + config = parse_config() + main(config) diff --git a/models/cv/classification/efficientnetv2_rw_t/ixrt/modify_batchsize.py b/models/cv/classification/efficientnetv2_rw_t/ixrt/modify_batchsize.py new file mode 100644 index 0000000000000000000000000000000000000000..689b7a972dcbfec77c185592ede16bb4f04fa4fd --- /dev/null +++ b/models/cv/classification/efficientnetv2_rw_t/ixrt/modify_batchsize.py @@ -0,0 +1,56 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import onnx +import argparse + +def change_input_dim(model, bsz): + batch_size = bsz + + # The following code changes the first dimension of every input to be batch_size + # Modify as appropriate ... note that this requires all inputs to + # have the same batch_size + inputs = model.graph.input + for input in inputs: + # Checks omitted.This assumes that all inputs are tensors and have a shape with first dim. + # Add checks as needed. + dim1 = input.type.tensor_type.shape.dim[0] + # update dim to be a symbolic value + if isinstance(batch_size, str): + # set dynamic batch size + dim1.dim_param = batch_size + elif (isinstance(batch_size, str) and batch_size.isdigit()) or isinstance(batch_size, int): + # set given batch size + dim1.dim_value = int(batch_size) + else: + # set batch size of 1 + dim1.dim_value = 1 + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--batch_size", type=int) + parser.add_argument("--origin_model", type=str) + parser.add_argument("--output_model", type=str) + args = parser.parse_args() + return args + +args = parse_args() +model = onnx.load(args.origin_model) +change_input_dim(model, args.batch_size) +onnx.save(model, args.output_model) + + + + + diff --git a/models/cv/classification/efficientnetv2_rw_t/ixrt/scripts/infer_efficientnetv2_rw_t_fp16_accuracy.sh b/models/cv/classification/efficientnetv2_rw_t/ixrt/scripts/infer_efficientnetv2_rw_t_fp16_accuracy.sh new file mode 100644 index 0000000000000000000000000000000000000000..86ecfac0a1285b9c03f477489eaa94fbc57951eb --- /dev/null +++ b/models/cv/classification/efficientnetv2_rw_t/ixrt/scripts/infer_efficientnetv2_rw_t_fp16_accuracy.sh @@ -0,0 +1,119 @@ +#!/bin/bash +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +EXIT_STATUS=0 +check_status() +{ + if ((${PIPESTATUS[0]} != 0));then + EXIT_STATUS=1 + fi +} + +# Run paraments +BSZ=32 +TGT=-1 +WARM_UP=0 +LOOP_COUNT=-1 +RUN_MODE=ACC +PRECISION=float16 + +# Update arguments +index=0 +options=$@ +arguments=($options) +for argument in $options +do + index=`expr $index + 1` + case $argument in + --bs) BSZ=${arguments[index]};; + --tgt) TGT=${arguments[index]};; + esac +done + +PROJ_DIR=${PROJ_DIR:-"."} +DATASETS_DIR="${DATASETS_DIR}" +CHECKPOINTS_DIR="${PROJ_DIR}" +RUN_DIR="${PROJ_DIR}" +CONFIG_DIR="${PROJ_DIR}/config/EFFICIENTNETV2_CONFIG" +source ${CONFIG_DIR} +ORIGINE_MODEL=${CHECKPOINTS_DIR}/${ORIGINE_MODEL} + +echo CHECKPOINTS_DIR : ${CHECKPOINTS_DIR} +echo DATASETS_DIR : ${DATASETS_DIR} +echo RUN_DIR : ${RUN_DIR} +echo CONFIG_DIR : ${CONFIG_DIR} +echo ====================== Model Info ====================== +echo Model Name : ${MODEL_NAME} +echo Onnx Path : ${ORIGINE_MODEL} + +step=0 +SIM_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_sim.onnx + +# Simplify Model +let step++ +echo; +echo [STEP ${step}] : Simplify Model +if [ -f ${SIM_MODEL} ];then + echo " "Simplify Model, ${SIM_MODEL} has been existed +else + python3 ${RUN_DIR}/simplify_model.py \ + --origin_model $ORIGINE_MODEL \ + --output_model ${SIM_MODEL} + echo " "Generate ${SIM_MODEL} +fi + +# Change Batchsize +let step++ +echo; +echo [STEP ${step}] : Change Batchsize +FINAL_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_${BSZ}.onnx +if [ -f $FINAL_MODEL ];then + echo " "Change Batchsize Skip, $FINAL_MODEL has been existed +else + python3 ${RUN_DIR}/modify_batchsize.py --batch_size ${BSZ} \ + --origin_model ${SIM_MODEL} --output_model ${FINAL_MODEL} + echo " "Generate ${FINAL_MODEL} +fi + +# Build Engine +let step++ +echo; +echo [STEP ${step}] : Build Engine +ENGINE_FILE=${CHECKPOINTS_DIR}/${MODEL_NAME}_${PRECISION}_bs${BSZ}.engine +if [ -f $ENGINE_FILE ];then + echo " "Build Engine Skip, $ENGINE_FILE has been existed +else + python3 ${RUN_DIR}/build_engine.py \ + --precision ${PRECISION} \ + --model ${FINAL_MODEL} \ + --engine ${ENGINE_FILE} + echo " "Generate Engine ${ENGINE_FILE} +fi + +# Inference +let step++ +echo; +echo [STEP ${step}] : Inference +python3 ${RUN_DIR}/inference.py \ + --engine_file=${ENGINE_FILE} \ + --datasets_dir=${DATASETS_DIR} \ + --imgsz=${IMGSIZE} \ + --warm_up=${WARM_UP} \ + --loop_count ${LOOP_COUNT} \ + --test_mode ${RUN_MODE} \ + --acc_target ${TGT} \ + --bsz ${BSZ}; check_status + +exit ${EXIT_STATUS} diff --git a/models/cv/classification/efficientnetv2_rw_t/ixrt/scripts/infer_efficientnetv2_rw_t_fp16_performance.sh b/models/cv/classification/efficientnetv2_rw_t/ixrt/scripts/infer_efficientnetv2_rw_t_fp16_performance.sh new file mode 100644 index 0000000000000000000000000000000000000000..597a7a64eaaf78f07d9c71e72064b94cbbbdb9ab --- /dev/null +++ b/models/cv/classification/efficientnetv2_rw_t/ixrt/scripts/infer_efficientnetv2_rw_t_fp16_performance.sh @@ -0,0 +1,119 @@ +#!/bin/bash +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +EXIT_STATUS=0 +check_status() +{ + if ((${PIPESTATUS[0]} != 0));then + EXIT_STATUS=1 + fi +} + +# Run paraments +BSZ=32 +TGT=-1 +WARM_UP=3 +LOOP_COUNT=20 +RUN_MODE=FPS +PRECISION=float16 + +# Update arguments +index=0 +options=$@ +arguments=($options) +for argument in $options +do + index=`expr $index + 1` + case $argument in + --bs) BSZ=${arguments[index]};; + --tgt) TGT=${arguments[index]};; + esac +done + +PROJ_DIR=${PROJ_DIR:-"."} +DATASETS_DIR="${DATASETS_DIR}" +CHECKPOINTS_DIR="${PROJ_DIR}" +RUN_DIR="${PROJ_DIR}" +CONFIG_DIR="${PROJ_DIR}/config/EFFICIENTNETV2_CONFIG" +source ${CONFIG_DIR} +ORIGINE_MODEL=${CHECKPOINTS_DIR}/${ORIGINE_MODEL} + +echo CHECKPOINTS_DIR : ${CHECKPOINTS_DIR} +echo DATASETS_DIR : ${DATASETS_DIR} +echo RUN_DIR : ${RUN_DIR} +echo CONFIG_DIR : ${CONFIG_DIR} +echo ====================== Model Info ====================== +echo Model Name : ${MODEL_NAME} +echo Onnx Path : ${ORIGINE_MODEL} + +step=0 +SIM_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_sim.onnx + +# Simplify Model +let step++ +echo; +echo [STEP ${step}] : Simplify Model +if [ -f ${SIM_MODEL} ];then + echo " "Simplify Model, ${SIM_MODEL} has been existed +else + python3 ${RUN_DIR}/simplify_model.py \ + --origin_model $ORIGINE_MODEL \ + --output_model ${SIM_MODEL} + echo " "Generate ${SIM_MODEL} +fi + +# Change Batchsize +let step++ +echo; +echo [STEP ${step}] : Change Batchsize +FINAL_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_${BSZ}.onnx +if [ -f $FINAL_MODEL ];then + echo " "Change Batchsize Skip, $FINAL_MODEL has been existed +else + python3 ${RUN_DIR}/modify_batchsize.py --batch_size ${BSZ} \ + --origin_model ${SIM_MODEL} --output_model ${FINAL_MODEL} + echo " "Generate ${FINAL_MODEL} +fi + +# Build Engine +let step++ +echo; +echo [STEP ${step}] : Build Engine +ENGINE_FILE=${CHECKPOINTS_DIR}/${MODEL_NAME}_${PRECISION}_bs${BSZ}.engine +if [ -f $ENGINE_FILE ];then + echo " "Build Engine Skip, $ENGINE_FILE has been existed +else + python3 ${RUN_DIR}/build_engine.py \ + --precision ${PRECISION} \ + --model ${FINAL_MODEL} \ + --engine ${ENGINE_FILE} + echo " "Generate Engine ${ENGINE_FILE} +fi + +# Inference +let step++ +echo; +echo [STEP ${step}] : Inference +python3 ${RUN_DIR}/inference.py \ + --engine_file=${ENGINE_FILE} \ + --datasets_dir=${DATASETS_DIR} \ + --imgsz=${IMGSIZE} \ + --warm_up=${WARM_UP} \ + --loop_count ${LOOP_COUNT} \ + --test_mode ${RUN_MODE} \ + --fps_target ${TGT} \ + --bsz ${BSZ}; check_status + +exit ${EXIT_STATUS} diff --git a/models/cv/classification/efficientnetv2_rw_t/ixrt/simplify_model.py b/models/cv/classification/efficientnetv2_rw_t/ixrt/simplify_model.py new file mode 100644 index 0000000000000000000000000000000000000000..9948a9fa083ff99ff88e556e96614b02cccaa965 --- /dev/null +++ b/models/cv/classification/efficientnetv2_rw_t/ixrt/simplify_model.py @@ -0,0 +1,40 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import onnx +import argparse +from onnxsim import simplify + +# Simplify +def simplify_model(args): + onnx_model = onnx.load(args.origin_model) + model_simp, check = simplify(onnx_model) + model_simp = onnx.shape_inference.infer_shapes(model_simp) + onnx.save(model_simp, args.output_model) + print(" Simplify onnx Done.") + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--origin_model", type=str) + parser.add_argument("--output_model", type=str) + parser.add_argument("--reshape", action="store_true") + args = parser.parse_args() + return args + +args = parse_args() +simplify_model(args) + + + + diff --git a/models/cv/detection/foveabox/ixrt/README.md b/models/cv/detection/foveabox/ixrt/README.md new file mode 100644 index 0000000000000000000000000000000000000000..d86a2c3961f6ac2da5700d345380ab4415331f6b --- /dev/null +++ b/models/cv/detection/foveabox/ixrt/README.md @@ -0,0 +1,67 @@ +# FoveaBox + +## Description + +FoveaBox is an advanced anchor-free object detection framework that enhances accuracy and flexibility by directly predicting the existence and bounding box coordinates of objects. Utilizing a Feature Pyramid Network (FPN), it adeptly handles targets of varying scales, particularly excelling with objects of arbitrary aspect ratios. FoveaBox also demonstrates robustness against image deformations. + +## Setup + +### Install + +```bash +# Install libGL +## CentOS +yum install -y mesa-libGL +## Ubuntu +apt install -y libgl1-mesa-dev + +pip3 install tqdm +pip3 install onnx +pip3 install onnxsim +pip3 install ultralytics +pip3 install pycocotools +pip3 install mmdeploy +pip3 install mmdet +pip3 install opencv-python==4.6.0.66 +``` + +### Download + +Pretrained model: + +Dataset: to download the validation dataset. + +### Model Conversion + +```bash +# export onnx model +python3 export.py --weight fovea_r50_fpn_4x4_1x_coco_20200219-ee4d5303.pth --cfg fovea_r50_fpn_4xb4-1x_coco.py --output foveabox.onnx + +# Use onnxsim optimize onnx model +onnxsim foveabox.onnx foveabox_opt.onnx +``` + +## Inference + +```bash +export DATASETS_DIR=/Path/to/coco/ +``` + +### FP16 + +```bash +# Accuracy +bash scripts/infer_foveabox_fp16_accuracy.sh +# Performance +bash scripts/infer_foveabox_fp16_performance.sh +``` + +## Results + +Model |BatchSize |Precision |FPS |IOU@0.5 |IOU@0.5:0.95 | +---------|-----------|----------|----------|----------|---------------| +FoveaBox | 32 | FP16 | 181.304 | 0.531 | 0.346 | + +## Reference + +mmdetection: diff --git a/models/cv/detection/foveabox/ixrt/build_engine.py b/models/cv/detection/foveabox/ixrt/build_engine.py new file mode 100644 index 0000000000000000000000000000000000000000..1bba10d0ee686cd6931161a249a6be887ea7cb24 --- /dev/null +++ b/models/cv/detection/foveabox/ixrt/build_engine.py @@ -0,0 +1,55 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import cv2 +import argparse +import numpy as np + +import torch +import tensorrt + +def main(config): + IXRT_LOGGER = tensorrt.Logger(tensorrt.Logger.WARNING) + builder = tensorrt.Builder(IXRT_LOGGER) + EXPLICIT_BATCH = 1 << (int)(tensorrt.NetworkDefinitionCreationFlag.EXPLICIT_BATCH) + network = builder.create_network(EXPLICIT_BATCH) + build_config = builder.create_builder_config() + parser = tensorrt.OnnxParser(network, IXRT_LOGGER) + parser.parse_from_file(config.model) + + precision = tensorrt.BuilderFlag.INT8 if config.precision == "int8" else tensorrt.BuilderFlag.FP16 + # print("precision : ", precision) + build_config.set_flag(precision) + + plan = builder.build_serialized_network(network, build_config) + engine_file_path = config.engine + with open(engine_file_path, "wb") as f: + f.write(plan) + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--model", type=str) + parser.add_argument("--precision", type=str, choices=["float16", "int8", "float32"], default="int8", + help="The precision of datatype") + # engine args + parser.add_argument("--engine", type=str, default=None) + + args = parser.parse_args() + return args + +if __name__ == "__main__": + args = parse_args() + main(args) \ No newline at end of file diff --git a/models/cv/detection/foveabox/ixrt/common.py b/models/cv/detection/foveabox/ixrt/common.py new file mode 100644 index 0000000000000000000000000000000000000000..ef92a6ba6291058d20f575edb09da35ebff3a937 --- /dev/null +++ b/models/cv/detection/foveabox/ixrt/common.py @@ -0,0 +1,69 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import cv2 +import glob +import torch +import tensorrt +import numpy as np +from cuda import cuda, cudart + +def create_engine_context(engine_path, logger): + with open(engine_path, "rb") as f: + runtime = tensorrt.Runtime(logger) + assert runtime + engine = runtime.deserialize_cuda_engine(f.read()) + assert engine + context = engine.create_execution_context() + assert context + + return engine, context + +def get_io_bindings(engine): + # Setup I/O bindings + inputs = [] + outputs = [] + allocations = [] + + for i in range(engine.num_bindings): + is_input = False + if engine.binding_is_input(i): + is_input = True + name = engine.get_binding_name(i) + dtype = engine.get_binding_dtype(i) + shape = engine.get_binding_shape(i) + if is_input: + batch_size = shape[0] + size = np.dtype(tensorrt.nptype(dtype)).itemsize + for s in shape: + size *= s + err, allocation = cudart.cudaMalloc(size) + assert err == cudart.cudaError_t.cudaSuccess + binding = { + "index": i, + "name": name, + "dtype": np.dtype(tensorrt.nptype(dtype)), + "shape": list(shape), + "allocation": allocation, + "nbytes": size, + } + print(f"binding {i}, name : {name} dtype : {np.dtype(tensorrt.nptype(dtype))} shape : {list(shape)}") + allocations.append(allocation) + if engine.binding_is_input(i): + inputs.append(binding) + else: + outputs.append(binding) + return inputs, outputs, allocations \ No newline at end of file diff --git a/models/cv/detection/foveabox/ixrt/config/FOVEABOX_CONFIG b/models/cv/detection/foveabox/ixrt/config/FOVEABOX_CONFIG new file mode 100644 index 0000000000000000000000000000000000000000..635c2b0de69ed23231edfa091badcffa3c5cc446 --- /dev/null +++ b/models/cv/detection/foveabox/ixrt/config/FOVEABOX_CONFIG @@ -0,0 +1,31 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +# BSZ : 构建engine以及推理时的batchsize +# IMGSIZE : 模型输入hw大小 +# RUN_MODE : [FPS, MAP] +# PRECISION : [float16, int8] +# MODEL_NAME : 生成onnx/engine的basename +# ORIGINE_MODEL : 原始onnx文件 +# COCO_GT : COCOEVAL标签文件 +# DATASET_DIR : 量化/推理数据集路径 +# CHECKPOINTS_DIR : 存放生成的onnx/engine路径 +# LAYER_FUSION : decoder部分走融合算子实现 0不融合 1融合 +# DECODER_FASTER : 有两种融合实现,faster版本速度快且可以直接对接gpu nms;另一种实现的输出和onnx保持一致. 1:faster +IMGSIZE=800 +MODEL_NAME=foveabox +ORIGINE_MODEL=foveabox_opt.onnx +DATA_PROCESS_TYPE=foveabox +MODEL_INPUT_NAMES=(input) diff --git a/models/cv/detection/foveabox/ixrt/deploy_default.py b/models/cv/detection/foveabox/ixrt/deploy_default.py new file mode 100644 index 0000000000000000000000000000000000000000..b8d8e43dc829456f0c2e46a7acfc3128757f945d --- /dev/null +++ b/models/cv/detection/foveabox/ixrt/deploy_default.py @@ -0,0 +1,41 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +onnx_config = dict( + type='onnx', + export_params=True, + keep_initializers_as_inputs=False, + opset_version=11, + save_file='end2end.onnx', + input_names=['input'], + output_names=['output'], + input_shape=None, + optimize=True) + +codebase_config = dict( + type='mmdet', + task='ObjectDetection', + model_type='end2end', + post_processing=dict( + score_threshold=0.05, + confidence_threshold=0.005, + iou_threshold=0.5, + max_output_boxes_per_class=200, + pre_top_k=5000, + keep_top_k=100, + background_label_id=-1, + )) + +backend_config = dict(type='onnxruntime') \ No newline at end of file diff --git a/models/cv/detection/foveabox/ixrt/export.py b/models/cv/detection/foveabox/ixrt/export.py new file mode 100644 index 0000000000000000000000000000000000000000..bceaba7801843feb5ef095f62a71a2f6e0074db4 --- /dev/null +++ b/models/cv/detection/foveabox/ixrt/export.py @@ -0,0 +1,74 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import argparse + +import torch +from mmdeploy.utils import load_config +from mmdeploy.apis import build_task_processor + +def parse_args(): + parser = argparse.ArgumentParser() + + parser.add_argument("--weight", + type=str, + required=True, + help="pytorch model weight.") + + parser.add_argument("--cfg", + type=str, + required=True, + help="model config file.") + + parser.add_argument("--output", + type=str, + required=True, + help="export onnx model path.") + + args = parser.parse_args() + return args + +def main(): + args = parse_args() + + deploy_cfg = 'deploy_default.py' + model_cfg = args.cfg + model_checkpoint = args.weight + + deploy_cfg, model_cfg = load_config(deploy_cfg, model_cfg) + + task_processor = build_task_processor(model_cfg, deploy_cfg, device='cpu') + + model = task_processor.build_pytorch_model(model_checkpoint) + + input_names = ['input'] + output_names = ['output'] + dynamic_axes = {'input': {0: '-1'}, 'output': {0: '-1'}} + dummy_input = torch.randn(1, 3, 800, 800) + + torch.onnx.export( + model, + dummy_input, + args.output, + input_names = input_names, + dynamic_axes = dynamic_axes, + output_names = output_names, + opset_version=13 + ) + + print("Export onnx model successfully! ") + +if __name__ == '__main__': + main() + diff --git a/models/cv/detection/foveabox/ixrt/fovea_r50_fpn_4xb4-1x_coco.py b/models/cv/detection/foveabox/ixrt/fovea_r50_fpn_4xb4-1x_coco.py new file mode 100644 index 0000000000000000000000000000000000000000..5be9e50af49009813315b0f7b9fc3221134d8ee1 --- /dev/null +++ b/models/cv/detection/foveabox/ixrt/fovea_r50_fpn_4xb4-1x_coco.py @@ -0,0 +1,287 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +auto_scale_lr = dict(base_batch_size=16, enable=False) +backend_args = None +data_root = 'data/coco/' +dataset_type = 'CocoDataset' +default_hooks = dict( + checkpoint=dict(interval=1, type='CheckpointHook'), + logger=dict(interval=50, type='LoggerHook'), + param_scheduler=dict(type='ParamSchedulerHook'), + sampler_seed=dict(type='DistSamplerSeedHook'), + timer=dict(type='IterTimerHook'), + visualization=dict(type='DetVisualizationHook')) +default_scope = 'mmdet' +env_cfg = dict( + cudnn_benchmark=False, + dist_cfg=dict(backend='nccl'), + mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0)) +load_from = None +log_level = 'INFO' +log_processor = dict(by_epoch=True, type='LogProcessor', window_size=50) +model = dict( + backbone=dict( + depth=50, + frozen_stages=1, + init_cfg=dict(checkpoint='torchvision://resnet50', type='Pretrained'), + norm_cfg=dict(requires_grad=True, type='BN'), + norm_eval=True, + num_stages=4, + out_indices=( + 0, + 1, + 2, + 3, + ), + style='pytorch', + type='ResNet'), + bbox_head=dict( + base_edge_list=[ + 16, + 32, + 64, + 128, + 256, + ], + feat_channels=256, + in_channels=256, + loss_bbox=dict(beta=0.11, loss_weight=1.0, type='SmoothL1Loss'), + loss_cls=dict( + alpha=0.4, + gamma=1.5, + loss_weight=1.0, + type='FocalLoss', + use_sigmoid=True), + num_classes=80, + scale_ranges=( + ( + 1, + 64, + ), + ( + 32, + 128, + ), + ( + 64, + 256, + ), + ( + 128, + 512, + ), + ( + 256, + 2048, + ), + ), + sigma=0.4, + stacked_convs=4, + strides=[ + 8, + 16, + 32, + 64, + 128, + ], + type='FoveaHead', + with_deform=False), + data_preprocessor=dict( + bgr_to_rgb=True, + mean=[ + 123.675, + 116.28, + 103.53, + ], + pad_size_divisor=32, + std=[ + 58.395, + 57.12, + 57.375, + ], + type='DetDataPreprocessor'), + neck=dict( + add_extra_convs='on_input', + in_channels=[ + 256, + 512, + 1024, + 2048, + ], + num_outs=5, + out_channels=256, + start_level=1, + type='FPN'), + test_cfg=dict( + max_per_img=100, + nms=dict(iou_threshold=0.5, type='nms'), + nms_pre=1000, + score_thr=0.05), + train_cfg=dict(), + type='FOVEA') +optim_wrapper = dict( + optimizer=dict(lr=0.01, momentum=0.9, type='SGD', weight_decay=0.0001), + type='OptimWrapper') +param_scheduler = [ + dict( + begin=0, by_epoch=False, end=500, start_factor=0.001, type='LinearLR'), + dict( + begin=0, + by_epoch=True, + end=12, + gamma=0.1, + milestones=[ + 8, + 11, + ], + type='MultiStepLR'), +] +resume = False +test_cfg = dict(type='TestLoop') +test_dataloader = dict( + batch_size=32, + dataset=dict( + ann_file='annotations/instances_val2017.json', + backend_args=None, + data_prefix=dict(img='images/val2017/'), + data_root='data/coco/', + pipeline=[ + dict(backend_args=None, type='LoadImageFromFile'), + dict(keep_ratio=True, scale=( + 800, + 800, + ), type='Resize'), + dict(type='LoadAnnotations', with_bbox=True), + dict( + meta_keys=( + 'img_id', + 'img_path', + 'ori_shape', + 'img_shape', + 'scale_factor', + ), + type='PackDetInputs'), + ], + test_mode=True, + type='CocoDataset'), + drop_last=False, + num_workers=2, + persistent_workers=True, + sampler=dict(shuffle=False, type='DefaultSampler')) +test_evaluator = dict( + ann_file='data/coco/annotations/instances_val2017.json', + backend_args=None, + format_only=False, + metric='bbox', + type='CocoMetric') +test_pipeline = [ + dict(backend_args=None, type='LoadImageFromFile'), + dict(keep_ratio=True, scale=( + 800, + 800, + ), type='Resize'), + dict(type='LoadAnnotations', with_bbox=True), + dict( + meta_keys=( + 'img_id', + 'img_path', + 'ori_shape', + 'img_shape', + 'scale_factor', + ), + type='PackDetInputs'), +] +train_cfg = dict(max_epochs=12, type='EpochBasedTrainLoop', val_interval=1) +train_dataloader = dict( + batch_sampler=dict(type='AspectRatioBatchSampler'), + batch_size=4, + dataset=dict( + ann_file='annotations/instances_train2017.json', + backend_args=None, + data_prefix=dict(img='train2017/'), + data_root='data/coco/', + filter_cfg=dict(filter_empty_gt=True, min_size=32), + pipeline=[ + dict(backend_args=None, type='LoadImageFromFile'), + dict(type='LoadAnnotations', with_bbox=True), + dict(keep_ratio=True, scale=( + 800, + 800, + ), type='Resize'), + dict(prob=0.5, type='RandomFlip'), + dict(type='PackDetInputs'), + ], + type='CocoDataset'), + num_workers=4, + persistent_workers=True, + sampler=dict(shuffle=True, type='DefaultSampler')) +train_pipeline = [ + dict(backend_args=None, type='LoadImageFromFile'), + dict(type='LoadAnnotations', with_bbox=True), + dict(keep_ratio=True, scale=( + 800, + 800, + ), type='Resize'), + dict(prob=0.5, type='RandomFlip'), + dict(type='PackDetInputs'), +] +val_cfg = dict(type='ValLoop') +val_dataloader = dict( + batch_size=1, + dataset=dict( + ann_file='annotations/instances_val2017.json', + backend_args=None, + data_prefix=dict(img='val2017/'), + data_root='data/coco/', + pipeline=[ + dict(backend_args=None, type='LoadImageFromFile'), + dict(keep_ratio=True, scale=( + 800, + 800, + ), type='Resize'), + dict(type='LoadAnnotations', with_bbox=True), + dict( + meta_keys=( + 'img_id', + 'img_path', + 'ori_shape', + 'img_shape', + 'scale_factor', + ), + type='PackDetInputs'), + ], + test_mode=True, + type='CocoDataset'), + drop_last=False, + num_workers=2, + persistent_workers=True, + sampler=dict(shuffle=False, type='DefaultSampler')) +val_evaluator = dict( + ann_file='data/coco/annotations/instances_val2017.json', + backend_args=None, + format_only=False, + metric='bbox', + type='CocoMetric') +vis_backends = [ + dict(type='LocalVisBackend'), +] +visualizer = dict( + name='visualizer', + type='DetLocalVisualizer', + vis_backends=[ + dict(type='LocalVisBackend'), + ]) +work_dir = './' diff --git a/models/cv/detection/foveabox/ixrt/inference.py b/models/cv/detection/foveabox/ixrt/inference.py new file mode 100644 index 0000000000000000000000000000000000000000..bcfe0e22d465482280149396b099fdd95e6681f7 --- /dev/null +++ b/models/cv/detection/foveabox/ixrt/inference.py @@ -0,0 +1,193 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import argparse +import torch +import torchvision +import numpy as np +import time +from tqdm import tqdm +from mmdet.registry import RUNNERS +from mmengine.config import Config + +import tensorrt +from common import create_engine_context, get_io_bindings +from cuda import cuda, cudart + +def parse_args(): + parser = argparse.ArgumentParser() + + parser.add_argument("--engine_file", + type=str, + required=True, + help="igie engine path.") + + parser.add_argument("--batchsize", + type=int, + required=True, + help="inference batch size.") + + parser.add_argument("--datasets", + type=str, + required=True, + help="datasets path.") + + parser.add_argument("--input_name", + type=str, + required=True, + help="input name of the model.") + + parser.add_argument("--warmup", + type=int, + default=3, + help="number of warmup before test.") + + parser.add_argument("--acc_target", + type=float, + default=None, + help="Model inference Accuracy target.") + + parser.add_argument("--fps_target", + type=float, + default=0.0, + help="Model inference FPS target.") + + parser.add_argument("--perf_only", + type=bool, + default=False, + help="Run performance test only") + + args = parser.parse_args() + + return args + +def main(): + args = parse_args() + + batch_size = args.batchsize + + host_mem = tensorrt.IHostMemory + logger = tensorrt.Logger(tensorrt.Logger.ERROR) + + # Load Engine && I/O bindings + engine, context = create_engine_context(args.engine_file, logger) + inputs, outputs, allocations = get_io_bindings(engine) + + # just run perf test + if args.perf_only: + torch.cuda.synchronize() + start_time = time.time() + + for i in range(100): + context.execute_v2(allocations) + + torch.cuda.synchronize() + end_time = time.time() + forward_time = end_time - start_time + + num_samples = 50000 + if 100 * batch_size < num_samples: + num_samples = 100 * batch_size + fps = num_samples / forward_time + + print("FPS : ", fps) + print(f"Performance Check : Test {fps} >= target {args.fps_target}") + if fps >= args.fps_target: + print("pass!") + exit() + else: + print("failed!") + exit(1) + else: + # warm up + print("\nWarm Start.") + for i in range(args.warmup): + context.execute_v2(allocations) + print("Warm Done.") + + # runner config + cfg = Config.fromfile("fovea_r50_fpn_4xb4-1x_coco.py") + + cfg.work_dir = "./workspace" + cfg['test_dataloader']['batch_size'] = batch_size + cfg['test_dataloader']['dataset']['data_root'] = args.datasets + cfg['test_dataloader']['dataset']['data_prefix']['img'] = 'images/val2017/' + cfg['test_evaluator']['ann_file'] = os.path.join(args.datasets, 'annotations/instances_val2017.json') + cfg['log_level'] = 'ERROR' + + # build runner + runner = RUNNERS.build(cfg) + + for input_data in tqdm(runner.test_dataloader): + + input_data = runner.model.data_preprocessor(input_data, False) + image = input_data['inputs'].cpu().numpy() + + pad_batch = len(image) != batch_size + if pad_batch: + origin_size = len(image) + image = np.resize(image, (batch_size, *image.shape[1:])) + + batch_data = image.astype(inputs[0]["dtype"]) + batch_data = np.ascontiguousarray(image) + + (err,) = cudart.cudaMemcpy( + inputs[0]["allocation"], + batch_data, + batch_data.nbytes, + cudart.cudaMemcpyKind.cudaMemcpyHostToDevice, + ) + assert err == cudart.cudaError_t.cudaSuccess + + context.execute_v2(allocations) + + cls_score = [] + box_reg = [] + + for i in range(len(outputs)): + output = np.zeros(outputs[i]["shape"], outputs[i]["dtype"]) + (err,) = cudart.cudaMemcpy( + output, + outputs[i]["allocation"], + outputs[i]["nbytes"], + cudart.cudaMemcpyKind.cudaMemcpyDeviceToHost, + ) + assert err == cudart.cudaError_t.cudaSuccess + + if pad_batch: + output = output[:origin_size] + + output = torch.from_numpy(output) + + if output.shape[1] == 80: + cls_score.append(output) + elif output.shape[1] == 4: + box_reg.append(output) + + batch_img_metas = [ + data_samples.metainfo for data_samples in input_data['data_samples'] + ] + + results_list = runner.model.bbox_head.predict_by_feat(cls_score, box_reg, batch_img_metas=batch_img_metas, rescale=True) + + batch_data_samples = runner.model.add_pred_to_datasample(input_data['data_samples'], results_list) + + runner.test_evaluator.process(data_samples=batch_data_samples, data_batch=input_data) + + metrics = runner.test_evaluator.evaluate(len(runner.test_dataloader.dataset)) + +if __name__ == "__main__": + main() \ No newline at end of file diff --git a/models/cv/detection/foveabox/ixrt/modify_batchsize.py b/models/cv/detection/foveabox/ixrt/modify_batchsize.py new file mode 100644 index 0000000000000000000000000000000000000000..3a88c1603bd6f457fd4965257627dc29edcda4d1 --- /dev/null +++ b/models/cv/detection/foveabox/ixrt/modify_batchsize.py @@ -0,0 +1,52 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import onnx +import argparse + +def change_input_dim(model, bsz): + batch_size = bsz + + # The following code changes the first dimension of every input to be batch_size + # Modify as appropriate ... note that this requires all inputs to + # have the same batch_size + inputs = model.graph.input + for input in inputs: + # Checks omitted.This assumes that all inputs are tensors and have a shape with first dim. + # Add checks as needed. + dim1 = input.type.tensor_type.shape.dim[0] + # update dim to be a symbolic value + if isinstance(batch_size, str): + # set dynamic batch size + dim1.dim_param = batch_size + elif (isinstance(batch_size, str) and batch_size.isdigit()) or isinstance(batch_size, int): + # set given batch size + dim1.dim_value = int(batch_size) + else: + # set batch size of 1 + dim1.dim_value = 1 + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--batch_size", type=int) + parser.add_argument("--origin_model", type=str) + parser.add_argument("--output_model", type=str) + args = parser.parse_args() + return args + +args = parse_args() +model = onnx.load(args.origin_model) +change_input_dim(model, args.batch_size) +onnx.save(model, args.output_model) \ No newline at end of file diff --git a/models/cv/detection/foveabox/ixrt/scripts/infer_foveabox_fp16_accuracy.sh b/models/cv/detection/foveabox/ixrt/scripts/infer_foveabox_fp16_accuracy.sh new file mode 100644 index 0000000000000000000000000000000000000000..4e73f22b7f1ab492676a7e6c0ebd36058b4fff2c --- /dev/null +++ b/models/cv/detection/foveabox/ixrt/scripts/infer_foveabox_fp16_accuracy.sh @@ -0,0 +1,123 @@ +#!/bin/bash +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +EXIT_STATUS=0 +check_status() +{ + if ((${PIPESTATUS[0]} != 0));then + EXIT_STATUS=1 + fi +} + +# Run paraments +BSZ=32 +WARM_UP=-1 +TGT=-1 +LOOP_COUNT=-1 +RUN_MODE=MAP +PRECISION=float16 + +# Update arguments +index=0 +options=$@ +arguments=($options) +for argument in $options +do + index=`expr $index + 1` + case $argument in + --bs) BSZ=${arguments[index]};; + --tgt) TGT=${arguments[index]};; + esac +done + +PROJ_DIR=${PROJ_DIR:-"."} +DATASETS_DIR="${DATASETS_DIR}" +CHECKPOINTS_DIR="${PROJ_DIR}" +RUN_DIR="${PROJ_DIR}" +CONFIG_DIR="${PROJ_DIR}/config/FOVEABOX_CONFIG" +source ${CONFIG_DIR} +ORIGINE_MODEL=${CHECKPOINTS_DIR}/${ORIGINE_MODEL} + +echo CHECKPOINTS_DIR : ${CHECKPOINTS_DIR} +echo DATASETS_DIR : ${DATASETS_DIR} +echo RUN_DIR : ${RUN_DIR} +echo CONFIG_DIR : ${CONFIG_DIR} +echo ====================== Model Info ====================== +echo Model Name : ${MODEL_NAME} +echo Onnx Path : ${ORIGINE_MODEL} + +step=0 +faster=0 +CURRENT_MODEL=${ORIGINE_MODEL} +if [[ ${LAYER_FUSION} == 1 && ${DECODER_FASTER} == 1 ]];then + faster=1 +fi + +# Simplify Model +let step++ +echo [STEP ${step}] : Simplify Model +SIM_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_sim.onnx +if [ -f ${SIM_MODEL} ];then + echo " "Simplify Model skip, ${SIM_MODEL} has been existed +else + python3 ${RUN_DIR}/simplify_model.py \ + --origin_model ${CURRENT_MODEL} \ + --output_model ${SIM_MODEL} + echo " "Generate ${SIM_MODEL} +fi +CURRENT_MODEL=${SIM_MODEL} + +# Change Batchsize +let step++ +echo; +echo [STEP ${step}] : Change Batchsize +FINAL_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_bs${BSZ}.onnx +if [ -f $FINAL_MODEL ];then + echo " "Change Batchsize Skip, $FINAL_MODEL has been existed +else + python3 ${RUN_DIR}/modify_batchsize.py \ + --batch_size ${BSZ} \ + --origin_model ${CURRENT_MODEL} \ + --output_model ${FINAL_MODEL} + echo " "Generate ${FINAL_MODEL} +fi +CURRENT_MODEL=${FINAL_MODEL} + +# Build Engine +let step++ +echo; +echo [STEP ${step}] : Build Engine +ENGINE_FILE=${CHECKPOINTS_DIR}/${MODEL_NAME}_${PRECISION}_bs${BSZ}.engine +if [ -f $ENGINE_FILE ];then + echo " "Build Engine Skip, $ENGINE_FILE has been existed +else + python3 ${RUN_DIR}/build_engine.py \ + --precision ${PRECISION} \ + --model ${CURRENT_MODEL} \ + --engine ${ENGINE_FILE} + echo " "Generate Engine ${ENGINE_FILE} +fi + +# Inference +let step++ +echo; +echo [STEP ${step}] : Inference +python3 ${RUN_DIR}/inference.py \ + --engine ${ENGINE_FILE} \ + --batchsize ${BSZ} \ + --input_name input \ + --datasets ${DATASETS_DIR}; check_status +exit ${EXIT_STATUS} \ No newline at end of file diff --git a/models/cv/detection/foveabox/ixrt/scripts/infer_foveabox_fp16_performance.sh b/models/cv/detection/foveabox/ixrt/scripts/infer_foveabox_fp16_performance.sh new file mode 100644 index 0000000000000000000000000000000000000000..eb96bcb158443a9605be5a641032c2661050e402 --- /dev/null +++ b/models/cv/detection/foveabox/ixrt/scripts/infer_foveabox_fp16_performance.sh @@ -0,0 +1,124 @@ +#!/bin/bash +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +EXIT_STATUS=0 +check_status() +{ + if ((${PIPESTATUS[0]} != 0));then + EXIT_STATUS=1 + fi +} + +# Run paraments +BSZ=32 +WARM_UP=-1 +TGT=-1 +LOOP_COUNT=-1 +RUN_MODE=FPS +PRECISION=float16 + +# Update arguments +index=0 +options=$@ +arguments=($options) +for argument in $options +do + index=`expr $index + 1` + case $argument in + --bs) BSZ=${arguments[index]};; + --tgt) TGT=${arguments[index]};; + esac +done + +PROJ_DIR=${PROJ_DIR:-"."} +DATASETS_DIR="${DATASETS_DIR}" +CHECKPOINTS_DIR="${PROJ_DIR}" +RUN_DIR="${PROJ_DIR}" +CONFIG_DIR="${PROJ_DIR}/config/FOVEABOX_CONFIG" +source ${CONFIG_DIR} +ORIGINE_MODEL=${CHECKPOINTS_DIR}/${ORIGINE_MODEL} + +echo CHECKPOINTS_DIR : ${CHECKPOINTS_DIR} +echo DATASETS_DIR : ${DATASETS_DIR} +echo RUN_DIR : ${RUN_DIR} +echo CONFIG_DIR : ${CONFIG_DIR} +echo ====================== Model Info ====================== +echo Model Name : ${MODEL_NAME} +echo Onnx Path : ${ORIGINE_MODEL} + +step=0 +faster=0 +CURRENT_MODEL=${ORIGINE_MODEL} +if [[ ${LAYER_FUSION} == 1 && ${DECODER_FASTER} == 1 ]];then + faster=1 +fi + +# Simplify Model +let step++ +echo [STEP ${step}] : Simplify Model +SIM_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_sim.onnx +if [ -f ${SIM_MODEL} ];then + echo " "Simplify Model skip, ${SIM_MODEL} has been existed +else + python3 ${RUN_DIR}/simplify_model.py \ + --origin_model ${CURRENT_MODEL} \ + --output_model ${SIM_MODEL} + echo " "Generate ${SIM_MODEL} +fi +CURRENT_MODEL=${SIM_MODEL} + +# Change Batchsize +let step++ +echo; +echo [STEP ${step}] : Change Batchsize +FINAL_MODEL=${CHECKPOINTS_DIR}/${MODEL_NAME}_bs${BSZ}.onnx +if [ -f $FINAL_MODEL ];then + echo " "Change Batchsize Skip, $FINAL_MODEL has been existed +else + python3 ${RUN_DIR}/modify_batchsize.py \ + --batch_size ${BSZ} \ + --origin_model ${CURRENT_MODEL} \ + --output_model ${FINAL_MODEL} + echo " "Generate ${FINAL_MODEL} +fi +CURRENT_MODEL=${FINAL_MODEL} + +# Build Engine +let step++ +echo; +echo [STEP ${step}] : Build Engine +ENGINE_FILE=${CHECKPOINTS_DIR}/${MODEL_NAME}_${PRECISION}_bs${BSZ}.engine +if [ -f $ENGINE_FILE ];then + echo " "Build Engine Skip, $ENGINE_FILE has been existed +else + python3 ${RUN_DIR}/build_engine.py \ + --precision ${PRECISION} \ + --model ${CURRENT_MODEL} \ + --engine ${ENGINE_FILE} + echo " "Generate Engine ${ENGINE_FILE} +fi + +# Inference +let step++ +echo; +echo [STEP ${step}] : Inference +python3 ${RUN_DIR}/inference.py \ + --engine ${ENGINE_FILE} \ + --batchsize ${BSZ} \ + --input_name input \ + --perf_only True \ + --datasets ${DATASETS_DIR}; check_status +exit ${EXIT_STATUS} \ No newline at end of file diff --git a/models/cv/detection/foveabox/ixrt/simplify_model.py b/models/cv/detection/foveabox/ixrt/simplify_model.py new file mode 100644 index 0000000000000000000000000000000000000000..1400fd81ddb4b3fae1b20d0fd35082a692f5d292 --- /dev/null +++ b/models/cv/detection/foveabox/ixrt/simplify_model.py @@ -0,0 +1,36 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import onnx +import argparse +from onnxsim import simplify + +# Simplify +def simplify_model(args): + onnx_model = onnx.load(args.origin_model) + model_simp, check = simplify(onnx_model) + model_simp = onnx.shape_inference.infer_shapes(model_simp) + onnx.save(model_simp, args.output_model) + print(" Simplify onnx Done.") + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--origin_model", type=str) + parser.add_argument("--output_model", type=str) + args = parser.parse_args() + return args + +args = parse_args() +simplify_model(args) \ No newline at end of file diff --git a/models/cv/detection/fsaf/ixrt/README.md b/models/cv/detection/fsaf/ixrt/README.md new file mode 100644 index 0000000000000000000000000000000000000000..5d75a2e6e3d0df468786d9f99a416b331c59d8c5 --- /dev/null +++ b/models/cv/detection/fsaf/ixrt/README.md @@ -0,0 +1,64 @@ +# FSAF + +## Description + +The FSAF (Feature Selective Anchor-Free) module is an innovative component for single-shot object detection that enhances performance through online feature selection and anchor-free branches. The FSAF module dynamically selects the most suitable feature level for each object instance, rather than relying on traditional anchor-based heuristic methods. This improvement significantly boosts the accuracy of object detection, especially for small targets and in complex scenes. Moreover, compared to existing anchor-based detectors, the FSAF module maintains high efficiency while adding negligible additional inference overhead. + +## Setup + +### Install + +```bash +# Install libGL +## CentOS +yum install -y mesa-libGL +## Ubuntu +apt install -y libgl1-mesa-dev + +pip3 install -r requirements.txt +``` + +### Download + +Pretrained model: + +Dataset: to download the validation dataset. + +```bash +wget https://download.openmmlab.com/mmdetection/v2.0/fsaf/fsaf_r50_fpn_1x_coco/fsaf_r50_fpn_1x_coco-94ccc51f.pth +``` + +### Model Conversion + +```bash +# export onnx model +python3 export.py --weight fsaf_r50_fpn_1x_coco-94ccc51f.pth --cfg fsaf_r50_fpn_1x_coco.py --output fsaf.onnx + +# use onnxsim optimize onnx model +onnxsim fsaf.onnx fsaf_opt.onnx +``` + +## Inference + +```bash +export DATASETS_DIR=/Path/to/coco/ +``` + +### FP16 + +```bash +# Accuracy +bash scripts/infer_fsaf_fp16_accuracy.sh +# Performance +bash scripts/infer_fsaf_fp16_performance.sh +``` + +## Results + +Model |BatchSize |Precision |FPS |IOU@0.5 |IOU@0.5:0.95 | +-------|-----------|----------|----------|----------|---------------| +FSAF | 32 | FP16 | 133.85 | 0.530 | 0.345 | + +## Reference + +mmdetection: diff --git a/models/cv/detection/fsaf/ixrt/build_engine.py b/models/cv/detection/fsaf/ixrt/build_engine.py new file mode 100644 index 0000000000000000000000000000000000000000..d7ad10e5e6da847867df40aa68c5a4f8710753ea --- /dev/null +++ b/models/cv/detection/fsaf/ixrt/build_engine.py @@ -0,0 +1,62 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import cv2 +import argparse +import numpy as np + +import torch +import tensorrt +from tensorrt import Dims + +def main(config): + IXRT_LOGGER = tensorrt.Logger(tensorrt.Logger.WARNING) + builder = tensorrt.Builder(IXRT_LOGGER) + EXPLICIT_BATCH = 1 << (int)(tensorrt.NetworkDefinitionCreationFlag.EXPLICIT_BATCH) + network = builder.create_network(EXPLICIT_BATCH) + build_config = builder.create_builder_config() + profile = builder.create_optimization_profile() + profile.set_shape("input", Dims([32, 3, 800, 800]), Dims([32, 3, 800, 800]), Dims([32, 3, 800, 800])) + build_config.add_optimization_profile(profile) + parser = tensorrt.OnnxParser(network, IXRT_LOGGER) + parser.parse_from_file(config.model) + + precision = tensorrt.BuilderFlag.INT8 if config.precision == "int8" else tensorrt.BuilderFlag.FP16 + # print("precision : ", precision) + build_config.set_flag(precision) + num_inputs = network.num_inputs + + for i in range(num_inputs): + input_tensor = network.get_input(i) + input_tensor.shape = Dims([32, 3, 800, 800]) + + plan = builder.build_serialized_network(network, build_config) + engine_file_path = config.engine + with open(engine_file_path, "wb") as f: + f.write(plan) + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--model", type=str) + parser.add_argument("--precision", type=str, choices=["float16", "int8", "float32"], default="float16", + help="The precision of datatype") + parser.add_argument("--engine", type=str, default=None) + args = parser.parse_args() + return args + +if __name__ == "__main__": + args = parse_args() + main(args) \ No newline at end of file diff --git a/models/cv/detection/fsaf/ixrt/common.py b/models/cv/detection/fsaf/ixrt/common.py new file mode 100644 index 0000000000000000000000000000000000000000..ef92a6ba6291058d20f575edb09da35ebff3a937 --- /dev/null +++ b/models/cv/detection/fsaf/ixrt/common.py @@ -0,0 +1,69 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import cv2 +import glob +import torch +import tensorrt +import numpy as np +from cuda import cuda, cudart + +def create_engine_context(engine_path, logger): + with open(engine_path, "rb") as f: + runtime = tensorrt.Runtime(logger) + assert runtime + engine = runtime.deserialize_cuda_engine(f.read()) + assert engine + context = engine.create_execution_context() + assert context + + return engine, context + +def get_io_bindings(engine): + # Setup I/O bindings + inputs = [] + outputs = [] + allocations = [] + + for i in range(engine.num_bindings): + is_input = False + if engine.binding_is_input(i): + is_input = True + name = engine.get_binding_name(i) + dtype = engine.get_binding_dtype(i) + shape = engine.get_binding_shape(i) + if is_input: + batch_size = shape[0] + size = np.dtype(tensorrt.nptype(dtype)).itemsize + for s in shape: + size *= s + err, allocation = cudart.cudaMalloc(size) + assert err == cudart.cudaError_t.cudaSuccess + binding = { + "index": i, + "name": name, + "dtype": np.dtype(tensorrt.nptype(dtype)), + "shape": list(shape), + "allocation": allocation, + "nbytes": size, + } + print(f"binding {i}, name : {name} dtype : {np.dtype(tensorrt.nptype(dtype))} shape : {list(shape)}") + allocations.append(allocation) + if engine.binding_is_input(i): + inputs.append(binding) + else: + outputs.append(binding) + return inputs, outputs, allocations \ No newline at end of file diff --git a/models/cv/detection/fsaf/ixrt/deploy_default.py b/models/cv/detection/fsaf/ixrt/deploy_default.py new file mode 100644 index 0000000000000000000000000000000000000000..b8d8e43dc829456f0c2e46a7acfc3128757f945d --- /dev/null +++ b/models/cv/detection/fsaf/ixrt/deploy_default.py @@ -0,0 +1,41 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +onnx_config = dict( + type='onnx', + export_params=True, + keep_initializers_as_inputs=False, + opset_version=11, + save_file='end2end.onnx', + input_names=['input'], + output_names=['output'], + input_shape=None, + optimize=True) + +codebase_config = dict( + type='mmdet', + task='ObjectDetection', + model_type='end2end', + post_processing=dict( + score_threshold=0.05, + confidence_threshold=0.005, + iou_threshold=0.5, + max_output_boxes_per_class=200, + pre_top_k=5000, + keep_top_k=100, + background_label_id=-1, + )) + +backend_config = dict(type='onnxruntime') \ No newline at end of file diff --git a/models/cv/detection/fsaf/ixrt/export.py b/models/cv/detection/fsaf/ixrt/export.py new file mode 100644 index 0000000000000000000000000000000000000000..13573c9dff3d96be4ba59eaa8698d67fb1d50f13 --- /dev/null +++ b/models/cv/detection/fsaf/ixrt/export.py @@ -0,0 +1,72 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import argparse + +import torch +from mmdeploy.utils import load_config +from mmdeploy.apis import build_task_processor + +def parse_args(): + parser = argparse.ArgumentParser() + + parser.add_argument("--weight", + type=str, + required=True, + help="pytorch model weight.") + + parser.add_argument("--cfg", + type=str, + required=True, + help="model config file.") + + parser.add_argument("--output", + type=str, + required=True, + help="export onnx model path.") + + args = parser.parse_args() + return args + +def main(): + args = parse_args() + + deploy_cfg = 'deploy_default.py' + model_cfg = args.cfg + model_checkpoint = args.weight + + deploy_cfg, model_cfg = load_config(deploy_cfg, model_cfg) + + task_processor = build_task_processor(model_cfg, deploy_cfg, device='cpu') + + model = task_processor.build_pytorch_model(model_checkpoint) + + input_names = ['input'] + dynamic_axes = {'input': {0: '-1'}} + dummy_input = torch.randn(1, 3, 800, 800) + + torch.onnx.export( + model, + dummy_input, + args.output, + input_names = input_names, + dynamic_axes = dynamic_axes, + opset_version=13 + ) + + print("Export onnx model successfully! ") + +if __name__ == '__main__': + main() \ No newline at end of file diff --git a/models/cv/detection/fsaf/ixrt/fsaf_r50_fpn_1x_coco.py b/models/cv/detection/fsaf/ixrt/fsaf_r50_fpn_1x_coco.py new file mode 100644 index 0000000000000000000000000000000000000000..d511321f40a1a3ca786e807d6f3a88f60c8bb994 --- /dev/null +++ b/models/cv/detection/fsaf/ixrt/fsaf_r50_fpn_1x_coco.py @@ -0,0 +1,276 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +auto_scale_lr = dict(base_batch_size=16, enable=False) +backend_args = None +data_root = 'data/coco/' +dataset_type = 'CocoDataset' +default_hooks = dict( + checkpoint=dict(interval=1, type='CheckpointHook'), + logger=dict(interval=50, type='LoggerHook'), + param_scheduler=dict(type='ParamSchedulerHook'), + sampler_seed=dict(type='DistSamplerSeedHook'), + timer=dict(type='IterTimerHook'), + visualization=dict(type='DetVisualizationHook')) +default_scope = 'mmdet' +env_cfg = dict( + cudnn_benchmark=False, + dist_cfg=dict(backend='nccl'), + mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0)) +load_from = None +log_level = 'ERROR' +log_processor = dict(by_epoch=True, type='LogProcessor', window_size=50) +model = dict( + backbone=dict( + depth=50, + frozen_stages=1, + init_cfg=dict(checkpoint='torchvision://resnet50', type='Pretrained'), + norm_cfg=dict(requires_grad=True, type='BN'), + norm_eval=True, + num_stages=4, + out_indices=( + 0, + 1, + 2, + 3, + ), + style='pytorch', + type='ResNet'), + bbox_head=dict( + anchor_generator=dict( + octave_base_scale=1, + ratios=[ + 1.0, + ], + scales_per_octave=1, + strides=[ + 8, + 16, + 32, + 64, + 128, + ], + type='AnchorGenerator'), + bbox_coder=dict(normalizer=4.0, type='TBLRBBoxCoder'), + feat_channels=256, + in_channels=256, + loss_bbox=dict( + eps=1e-06, loss_weight=1.0, reduction='none', type='IoULoss'), + loss_cls=dict( + alpha=0.25, + gamma=2.0, + loss_weight=1.0, + reduction='none', + type='FocalLoss', + use_sigmoid=True), + num_classes=80, + reg_decoded_bbox=True, + stacked_convs=4, + type='FSAFHead'), + data_preprocessor=dict( + bgr_to_rgb=True, + mean=[ + 123.675, + 116.28, + 103.53, + ], + pad_size_divisor=32, + std=[ + 58.395, + 57.12, + 57.375, + ], + type='DetDataPreprocessor'), + neck=dict( + add_extra_convs='on_input', + in_channels=[ + 256, + 512, + 1024, + 2048, + ], + num_outs=5, + out_channels=256, + start_level=1, + type='FPN'), + test_cfg=dict( + max_per_img=100, + min_bbox_size=0, + nms=dict(iou_threshold=0.5, type='nms'), + nms_pre=1000, + score_thr=0.05), + train_cfg=dict( + allowed_border=-1, + assigner=dict( + min_pos_iof=0.01, + neg_scale=0.2, + pos_scale=0.2, + type='CenterRegionAssigner'), + debug=False, + pos_weight=-1, + sampler=dict(type='PseudoSampler')), + type='FSAF') +optim_wrapper = dict( + optimizer=dict(lr=0.02, momentum=0.9, type='SGD', weight_decay=0.0001), + type='OptimWrapper') +param_scheduler = [ + dict( + begin=0, by_epoch=False, end=500, start_factor=0.001, type='LinearLR'), + dict( + begin=0, + by_epoch=True, + end=12, + gamma=0.1, + milestones=[ + 8, + 11, + ], + type='MultiStepLR'), +] +resume = False +test_cfg = dict(type='TestLoop') +test_dataloader = dict( + batch_size=32, + dataset=dict( + ann_file='annotations/instances_val2017.json', + backend_args=None, + data_prefix=dict(img='images/val2017/'), + data_root='data/coco/', + pipeline=[ + dict(backend_args=None, type='LoadImageFromFile'), + dict(keep_ratio=False, scale=( + 800, + 800, + ), type='Resize'), + dict(type='LoadAnnotations', with_bbox=True), + dict( + meta_keys=( + 'img_id', + 'img_path', + 'ori_shape', + 'img_shape', + 'scale_factor', + ), + type='PackDetInputs'), + ], + test_mode=True, + type='CocoDataset'), + drop_last=False, + num_workers=2, + persistent_workers=True, + sampler=dict(shuffle=False, type='DefaultSampler')) +test_evaluator = dict( + ann_file='data/coco/annotations/instances_val2017.json', + backend_args=None, + format_only=False, + metric='bbox', + type='CocoMetric') +test_pipeline = [ + dict(backend_args=None, type='LoadImageFromFile'), + dict(keep_ratio=True, scale=( + 1333, + 800, + ), type='Resize'), + dict(type='LoadAnnotations', with_bbox=True), + dict( + meta_keys=( + 'img_id', + 'img_path', + 'ori_shape', + 'img_shape', + 'scale_factor', + ), + type='PackDetInputs'), +] +train_cfg = dict(max_epochs=12, type='EpochBasedTrainLoop', val_interval=1) +train_dataloader = dict( + batch_sampler=dict(type='AspectRatioBatchSampler'), + batch_size=2, + dataset=dict( + ann_file='annotations/instances_train2017.json', + backend_args=None, + data_prefix=dict(img='train2017/'), + data_root='data/coco/', + filter_cfg=dict(filter_empty_gt=True, min_size=32), + pipeline=[ + dict(backend_args=None, type='LoadImageFromFile'), + dict(type='LoadAnnotations', with_bbox=True), + dict(keep_ratio=True, scale=( + 1333, + 800, + ), type='Resize'), + dict(prob=0.5, type='RandomFlip'), + dict(type='PackDetInputs'), + ], + type='CocoDataset'), + num_workers=2, + persistent_workers=True, + sampler=dict(shuffle=True, type='DefaultSampler')) +train_pipeline = [ + dict(backend_args=None, type='LoadImageFromFile'), + dict(type='LoadAnnotations', with_bbox=True), + dict(keep_ratio=True, scale=( + 1333, + 800, + ), type='Resize'), + dict(prob=0.5, type='RandomFlip'), + dict(type='PackDetInputs'), +] +val_cfg = dict(type='ValLoop') +val_dataloader = dict( + batch_size=1, + dataset=dict( + ann_file='annotations/instances_val2017.json', + backend_args=None, + data_prefix=dict(img='val2017/'), + data_root='data/coco/', + pipeline=[ + dict(backend_args=None, type='LoadImageFromFile'), + dict(keep_ratio=True, scale=( + 1333, + 800, + ), type='Resize'), + dict(type='LoadAnnotations', with_bbox=True), + dict( + meta_keys=( + 'img_id', + 'img_path', + 'ori_shape', + 'img_shape', + 'scale_factor', + ), + type='PackDetInputs'), + ], + test_mode=True, + type='CocoDataset'), + drop_last=False, + num_workers=2, + persistent_workers=True, + sampler=dict(shuffle=False, type='DefaultSampler')) +val_evaluator = dict( + ann_file='data/coco/annotations/instances_val2017.json', + backend_args=None, + format_only=False, + metric='bbox', + type='CocoMetric') +vis_backends = [ + dict(type='LocalVisBackend'), +] +visualizer = dict( + name='visualizer', + type='DetLocalVisualizer', + vis_backends=[ + dict(type='LocalVisBackend'), + ]) \ No newline at end of file diff --git a/models/cv/detection/fsaf/ixrt/inference.py b/models/cv/detection/fsaf/ixrt/inference.py new file mode 100644 index 0000000000000000000000000000000000000000..5d940cd5d656dea893acc1993992994a3c4de0a5 --- /dev/null +++ b/models/cv/detection/fsaf/ixrt/inference.py @@ -0,0 +1,187 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import time +import argparse +import tensorrt +import torch +import torchvision +import numpy as np +from tensorrt import Dims +from cuda import cuda, cudart +from tqdm import tqdm +from mmdet.registry import RUNNERS +from mmengine.config import Config + +from common import create_engine_context, get_io_bindings + +def parse_args(): + parser = argparse.ArgumentParser() + + parser.add_argument("--engine", + type=str, + required=True, + help="igie engine path.") + + parser.add_argument("--batchsize", + type=int, + required=True, + help="inference batch size.") + + parser.add_argument("--datasets", + type=str, + required=True, + help="datasets path.") + + parser.add_argument("--input_name", + type=str, + required=True, + help="input name of the model.") + + parser.add_argument("--warmup", + type=int, + default=3, + help="number of warmup before test.") + + parser.add_argument("--acc_target", + type=float, + default=None, + help="Model inference Accuracy target.") + + parser.add_argument("--fps_target", + type=float, + default=None, + help="Model inference FPS target.") + + parser.add_argument("--perf_only", + type=bool, + default=False, + help="Run performance test only") + + args = parser.parse_args() + + return args + +def main(): + args = parse_args() + + batch_size = args.batchsize + + host_mem = tensorrt.IHostMemory + logger = tensorrt.Logger(tensorrt.Logger.ERROR) + + # Load Engine && I/O bindings + engine, context = create_engine_context(args.engine, logger) + inputs, outputs, allocations = get_io_bindings(engine) + + # just run perf test + if args.perf_only: + torch.cuda.synchronize() + start_time = time.time() + + for i in range(10): + context.execute_v2(allocations) + + torch.cuda.synchronize() + end_time = time.time() + forward_time = end_time - start_time + num_samples = 10 * args.batchsize + fps = num_samples / forward_time + + print("FPS : ", fps) + print(f"Performance Check : Test {fps} >= target {args.fps_target}") + if fps >= args.fps_target: + print("pass!") + exit() + else: + print("failed!") + exit(1) + else: + # runner config + cfg = Config.fromfile("fsaf_r50_fpn_1x_coco.py") + + cfg.work_dir = "./workspace" + cfg['test_dataloader']['batch_size'] = batch_size + cfg['test_dataloader']['dataset']['data_root'] = args.datasets + cfg['test_dataloader']['dataset']['data_prefix']['img'] = 'images/val2017/' + cfg['test_evaluator']['ann_file'] = os.path.join(args.datasets, 'annotations/instances_val2017.json') + cfg['log_level'] = 'ERROR' + + # build runner + runner = RUNNERS.build(cfg) + + for data in tqdm(runner.test_dataloader): + cls_score = [] + box_reg = [] + + input_data = runner.model.data_preprocessor(data, False) + image = input_data['inputs'].cpu() + image = image.numpy().astype(inputs[0]["dtype"]) + pad_batch = len(image) != batch_size + + if pad_batch: + origin_size = len(image) + image = np.resize(image, (batch_size, *image.shape[1:])) + + image = np.ascontiguousarray(image) + + (err,) = cudart.cudaMemcpy( + inputs[0]["allocation"], + image, + image.nbytes, + cudart.cudaMemcpyKind.cudaMemcpyHostToDevice, + ) + assert err == cudart.cudaError_t.cudaSuccess + # cuda.memcpy_htod(inputs[0]["allocation"], batch_data) + context.execute_v2(allocations) + + for i in range(len(outputs)): + output = np.zeros(outputs[i]["shape"], outputs[i]["dtype"]) + (err,) = cudart.cudaMemcpy( + output, + outputs[i]["allocation"], + outputs[i]["nbytes"], + cudart.cudaMemcpyKind.cudaMemcpyDeviceToHost, + ) + assert err == cudart.cudaError_t.cudaSuccess + + if pad_batch: + output = output[:origin_size] + + output = torch.from_numpy(output) + + if output.shape[1] == 80: + cls_score.append(output) + elif output.shape[1] == 4: + box_reg.append(output) + + batch_img_metas = [ + data_samples.metainfo for data_samples in data['data_samples'] + ] + + preds = runner.model.bbox_head.predict_by_feat( + cls_score, box_reg, batch_img_metas=batch_img_metas, rescale=True + ) + + batch_data_samples = runner.model.add_pred_to_datasample(input_data['data_samples'], preds) + + runner.test_evaluator.process(data_samples=batch_data_samples, data_batch=data) + + metrics = runner.test_evaluator.evaluate(len(runner.test_dataloader.dataset)) + + +if __name__ == "__main__": + main() diff --git a/models/cv/detection/fsaf/ixrt/requirements.txt b/models/cv/detection/fsaf/ixrt/requirements.txt new file mode 100644 index 0000000000000000000000000000000000000000..a26706ef5402ca820ed6e4ab952d876ec768b4eb --- /dev/null +++ b/models/cv/detection/fsaf/ixrt/requirements.txt @@ -0,0 +1,6 @@ +onnx +tqdm +onnxsim +mmdet==3.3.0 +mmdeploy==1.3.1 +mmengine==0.10.4 diff --git a/models/cv/detection/fsaf/ixrt/scripts/infer_fsaf_fp16_accuracy.sh b/models/cv/detection/fsaf/ixrt/scripts/infer_fsaf_fp16_accuracy.sh new file mode 100644 index 0000000000000000000000000000000000000000..ed3132c67b47f419fb43822f92530b65d58a742d --- /dev/null +++ b/models/cv/detection/fsaf/ixrt/scripts/infer_fsaf_fp16_accuracy.sh @@ -0,0 +1,34 @@ +#!/bin/bash + +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +batchsize=32 +model_path="fsaf_opt.onnx" +datasets_path=${DATASETS_DIR} + +# build engine +python3 build_engine.py \ + --model ${model_path} \ + --precision float16 \ + --engine fsaf.engine + + +# inference +python3 inference.py \ + --engine fsaf.engine \ + --batchsize ${batchsize} \ + --input_name input \ + --datasets ${datasets_path} \ No newline at end of file diff --git a/models/cv/detection/fsaf/ixrt/scripts/infer_fsaf_fp16_performance.sh b/models/cv/detection/fsaf/ixrt/scripts/infer_fsaf_fp16_performance.sh new file mode 100644 index 0000000000000000000000000000000000000000..65fad0c7ff07cc92f1be605db91f9e0cbf849a8c --- /dev/null +++ b/models/cv/detection/fsaf/ixrt/scripts/infer_fsaf_fp16_performance.sh @@ -0,0 +1,36 @@ +#!/bin/bash + +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +batchsize=32 +model_path="fsaf_opt.onnx" +datasets_path=${DATASETS_DIR} + +# build engine +python3 build_engine.py \ + --model ${model_path} \ + --precision float16 \ + --engine fsaf.engine + + +# inference +python3 inference.py \ + --engine fsaf.engine \ + --batchsize ${batchsize} \ + --input_name input \ + --datasets ${datasets_path} \ + --perf_only True \ + --fps_target 130 \ No newline at end of file diff --git a/models/cv/detection/hrnet/ixrt/README.md b/models/cv/detection/hrnet/ixrt/README.md new file mode 100644 index 0000000000000000000000000000000000000000..cade087c4e00dac7b92db9943932facf97ba04e4 --- /dev/null +++ b/models/cv/detection/hrnet/ixrt/README.md @@ -0,0 +1,60 @@ +# HRNet + +## Description + +HRNet is an advanced deep learning architecture for human pose estimation, characterized by its maintenance of high-resolution representations throughout the entire network process, thereby avoiding the low-to-high resolution recovery step typical of traditional models. The network features parallel multi-resolution subnetworks and enriches feature representation through repeated multi-scale fusion, which enhances the accuracy of keypoint detection. Additionally, HRNet offers computational efficiency and has demonstrated superior performance over previous methods on several standard datasets. + +## Setup + +### Install + +```bash +# Install libGL +## CentOS +yum install -y mesa-libGL +## Ubuntu +apt install -y libgl1-mesa-dev + +pip3 install -r requirements.txt +``` + +### Download + +Pretrained model: + +Dataset: to download the validation dataset. + +### Model Conversion + +```bash +# export onnx model +python3 export.py --weight fcos_hrnetv2p_w18_gn-head_4x4_1x_coco_20201212_100710-4ad151de.pth --cfg fcos_hrnetv2p-w18-gn-head_4xb4-1x_coco.py --output hrnet.onnx + +# Use onnxsim optimize onnx model +onnxsim hrnet.onnx hrnet_opt.onnx +``` + +## Inference + +```bash +export DATASETS_DIR=/Path/to/coco/ +``` + +### FP16 + +```bash +# Accuracy +bash scripts/infer_hrnet_fp16_accuracy.sh +# Performance +bash scripts/infer_hrnet_fp16_performance.sh +``` + +## Results + +Model |BatchSize |Precision |FPS |IOU@0.5 |IOU@0.5:0.95 | +-------|-----------|----------|----------|----------|---------------| +HRNet | 32 | FP16 | 75.199 | 0.491 | 0.327 | + +## Reference + +mmdetection: diff --git a/models/cv/detection/hrnet/ixrt/build_engine.py b/models/cv/detection/hrnet/ixrt/build_engine.py new file mode 100644 index 0000000000000000000000000000000000000000..61834f4fe79fe2b9bb4cfc01561b01f177098f69 --- /dev/null +++ b/models/cv/detection/hrnet/ixrt/build_engine.py @@ -0,0 +1,61 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import cv2 +import argparse +import numpy as np + +import torch +import tensorrt +from tensorrt import Dims + +def main(config): + IXRT_LOGGER = tensorrt.Logger(tensorrt.Logger.WARNING) + builder = tensorrt.Builder(IXRT_LOGGER) + EXPLICIT_BATCH = 1 << (int)(tensorrt.NetworkDefinitionCreationFlag.EXPLICIT_BATCH) + network = builder.create_network(EXPLICIT_BATCH) + build_config = builder.create_builder_config() + profile = builder.create_optimization_profile() + profile.set_shape("input", Dims([32, 3, 800, 800]), Dims([32, 3, 800, 800]), Dims([32, 3, 800, 800])) + build_config.add_optimization_profile(profile) + parser = tensorrt.OnnxParser(network, IXRT_LOGGER) + parser.parse_from_file(config.model) + + precision = tensorrt.BuilderFlag.INT8 if config.precision == "int8" else tensorrt.BuilderFlag.FP16 + build_config.set_flag(precision) + num_inputs = network.num_inputs + + for i in range(num_inputs): + input_tensor = network.get_input(i) + input_tensor.shape = Dims([32, 3, 800, 800]) + + plan = builder.build_serialized_network(network, build_config) + engine_file_path = config.engine + with open(engine_file_path, "wb") as f: + f.write(plan) + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--model", type=str) + parser.add_argument("--precision", type=str, choices=["float16", "int8", "float32"], default="float16", + help="The precision of datatype") + parser.add_argument("--engine", type=str, default=None) + args = parser.parse_args() + return args + +if __name__ == "__main__": + args = parse_args() + main(args) \ No newline at end of file diff --git a/models/cv/detection/hrnet/ixrt/common.py b/models/cv/detection/hrnet/ixrt/common.py new file mode 100644 index 0000000000000000000000000000000000000000..ef92a6ba6291058d20f575edb09da35ebff3a937 --- /dev/null +++ b/models/cv/detection/hrnet/ixrt/common.py @@ -0,0 +1,69 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import cv2 +import glob +import torch +import tensorrt +import numpy as np +from cuda import cuda, cudart + +def create_engine_context(engine_path, logger): + with open(engine_path, "rb") as f: + runtime = tensorrt.Runtime(logger) + assert runtime + engine = runtime.deserialize_cuda_engine(f.read()) + assert engine + context = engine.create_execution_context() + assert context + + return engine, context + +def get_io_bindings(engine): + # Setup I/O bindings + inputs = [] + outputs = [] + allocations = [] + + for i in range(engine.num_bindings): + is_input = False + if engine.binding_is_input(i): + is_input = True + name = engine.get_binding_name(i) + dtype = engine.get_binding_dtype(i) + shape = engine.get_binding_shape(i) + if is_input: + batch_size = shape[0] + size = np.dtype(tensorrt.nptype(dtype)).itemsize + for s in shape: + size *= s + err, allocation = cudart.cudaMalloc(size) + assert err == cudart.cudaError_t.cudaSuccess + binding = { + "index": i, + "name": name, + "dtype": np.dtype(tensorrt.nptype(dtype)), + "shape": list(shape), + "allocation": allocation, + "nbytes": size, + } + print(f"binding {i}, name : {name} dtype : {np.dtype(tensorrt.nptype(dtype))} shape : {list(shape)}") + allocations.append(allocation) + if engine.binding_is_input(i): + inputs.append(binding) + else: + outputs.append(binding) + return inputs, outputs, allocations \ No newline at end of file diff --git a/models/cv/detection/hrnet/ixrt/deploy_default.py b/models/cv/detection/hrnet/ixrt/deploy_default.py new file mode 100644 index 0000000000000000000000000000000000000000..b8d8e43dc829456f0c2e46a7acfc3128757f945d --- /dev/null +++ b/models/cv/detection/hrnet/ixrt/deploy_default.py @@ -0,0 +1,41 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +onnx_config = dict( + type='onnx', + export_params=True, + keep_initializers_as_inputs=False, + opset_version=11, + save_file='end2end.onnx', + input_names=['input'], + output_names=['output'], + input_shape=None, + optimize=True) + +codebase_config = dict( + type='mmdet', + task='ObjectDetection', + model_type='end2end', + post_processing=dict( + score_threshold=0.05, + confidence_threshold=0.005, + iou_threshold=0.5, + max_output_boxes_per_class=200, + pre_top_k=5000, + keep_top_k=100, + background_label_id=-1, + )) + +backend_config = dict(type='onnxruntime') \ No newline at end of file diff --git a/models/cv/detection/hrnet/ixrt/export.py b/models/cv/detection/hrnet/ixrt/export.py new file mode 100644 index 0000000000000000000000000000000000000000..bceaba7801843feb5ef095f62a71a2f6e0074db4 --- /dev/null +++ b/models/cv/detection/hrnet/ixrt/export.py @@ -0,0 +1,74 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +import argparse + +import torch +from mmdeploy.utils import load_config +from mmdeploy.apis import build_task_processor + +def parse_args(): + parser = argparse.ArgumentParser() + + parser.add_argument("--weight", + type=str, + required=True, + help="pytorch model weight.") + + parser.add_argument("--cfg", + type=str, + required=True, + help="model config file.") + + parser.add_argument("--output", + type=str, + required=True, + help="export onnx model path.") + + args = parser.parse_args() + return args + +def main(): + args = parse_args() + + deploy_cfg = 'deploy_default.py' + model_cfg = args.cfg + model_checkpoint = args.weight + + deploy_cfg, model_cfg = load_config(deploy_cfg, model_cfg) + + task_processor = build_task_processor(model_cfg, deploy_cfg, device='cpu') + + model = task_processor.build_pytorch_model(model_checkpoint) + + input_names = ['input'] + output_names = ['output'] + dynamic_axes = {'input': {0: '-1'}, 'output': {0: '-1'}} + dummy_input = torch.randn(1, 3, 800, 800) + + torch.onnx.export( + model, + dummy_input, + args.output, + input_names = input_names, + dynamic_axes = dynamic_axes, + output_names = output_names, + opset_version=13 + ) + + print("Export onnx model successfully! ") + +if __name__ == '__main__': + main() + diff --git a/models/cv/detection/hrnet/ixrt/fcos_hrnetv2p-w18-gn-head_4xb4-1x_coco.py b/models/cv/detection/hrnet/ixrt/fcos_hrnetv2p-w18-gn-head_4xb4-1x_coco.py new file mode 100644 index 0000000000000000000000000000000000000000..faccb7732d7063f680cf023fce4fc3ece716583d --- /dev/null +++ b/models/cv/detection/hrnet/ixrt/fcos_hrnetv2p-w18-gn-head_4xb4-1x_coco.py @@ -0,0 +1,287 @@ +auto_scale_lr = dict(base_batch_size=16, enable=False) +backend_args = None +data_root = 'data/coco/' +dataset_type = 'CocoDataset' +default_hooks = dict( + checkpoint=dict(interval=1, type='CheckpointHook'), + logger=dict(interval=50, type='LoggerHook'), + param_scheduler=dict(type='ParamSchedulerHook'), + sampler_seed=dict(type='DistSamplerSeedHook'), + timer=dict(type='IterTimerHook'), + visualization=dict(type='DetVisualizationHook')) +default_scope = 'mmdet' +env_cfg = dict( + cudnn_benchmark=False, + dist_cfg=dict(backend='nccl'), + mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0)) +load_from = None +log_level = 'ERROR' +log_processor = dict(by_epoch=True, type='LogProcessor', window_size=50) +model = dict( + backbone=dict( + extra=dict( + stage1=dict( + block='BOTTLENECK', + num_blocks=(4, ), + num_branches=1, + num_channels=(64, ), + num_modules=1), + stage2=dict( + block='BASIC', + num_blocks=( + 4, + 4, + ), + num_branches=2, + num_channels=( + 18, + 36, + ), + num_modules=1), + stage3=dict( + block='BASIC', + num_blocks=( + 4, + 4, + 4, + ), + num_branches=3, + num_channels=( + 18, + 36, + 72, + ), + num_modules=4), + stage4=dict( + block='BASIC', + num_blocks=( + 4, + 4, + 4, + 4, + ), + num_branches=4, + num_channels=( + 18, + 36, + 72, + 144, + ), + num_modules=3)), + init_cfg=dict( + checkpoint='open-mmlab://msra/hrnetv2_w18', type='Pretrained'), + type='HRNet'), + bbox_head=dict( + feat_channels=256, + in_channels=256, + loss_bbox=dict(loss_weight=1.0, type='IoULoss'), + loss_centerness=dict( + loss_weight=1.0, type='CrossEntropyLoss', use_sigmoid=True), + loss_cls=dict( + alpha=0.25, + gamma=2.0, + loss_weight=1.0, + type='FocalLoss', + use_sigmoid=True), + num_classes=80, + stacked_convs=4, + strides=[ + 8, + 16, + 32, + 64, + 128, + ], + type='FCOSHead'), + data_preprocessor=dict( + bgr_to_rgb=False, + mean=[ + 103.53, + 116.28, + 123.675, + ], + pad_size_divisor=32, + std=[ + 57.375, + 57.12, + 58.395, + ], + type='DetDataPreprocessor'), + neck=dict( + in_channels=[ + 18, + 36, + 72, + 144, + ], + num_outs=5, + out_channels=256, + stride=2, + type='HRFPN'), + test_cfg=dict( + max_per_img=100, + min_bbox_size=0, + nms=dict(iou_threshold=0.5, type='nms'), + nms_pre=1000, + score_thr=0.05), + type='FCOS') +optim_wrapper = dict( + clip_grad=dict(max_norm=35, norm_type=2), + optimizer=dict(lr=0.01, momentum=0.9, type='SGD', weight_decay=0.0001), + paramwise_cfg=dict(bias_decay_mult=0.0, bias_lr_mult=2.0), + type='OptimWrapper') +param_scheduler = [ + dict( + begin=0, + by_epoch=False, + end=500, + factor=0.3333333333333333, + type='ConstantLR'), + dict( + begin=0, + by_epoch=True, + end=12, + gamma=0.1, + milestones=[ + 8, + 11, + ], + type='MultiStepLR'), +] +resume = False +test_cfg = dict(type='TestLoop') +test_dataloader = dict( + batch_size=32, + dataset=dict( + ann_file='annotations/instances_val2017.json', + backend_args=None, + data_prefix=dict(img='val2017/'), + data_root='/home/xinchi.tian/ixrt-modelzoo/data/datasets/coco2017', + pipeline=[ + dict(backend_args=None, type='LoadImageFromFile'), + dict(keep_ratio=True, scale=( + 800, + 800, + ), type='Resize'), + dict(type='LoadAnnotations', with_bbox=True), + dict( + meta_keys=( + 'img_id', + 'img_path', + 'ori_shape', + 'img_shape', + 'scale_factor', + ), + type='PackDetInputs'), + ], + test_mode=True, + type='CocoDataset'), + drop_last=False, + num_workers=2, + persistent_workers=True, + sampler=dict(shuffle=False, type='DefaultSampler')) +test_evaluator = dict( + ann_file= + '/home/xinchi.tian/ixrt-modelzoo/data/datasets/coco2017/annotations/instances_val2017.json', + backend_args=None, + format_only=False, + metric='bbox', + type='CocoMetric') +test_pipeline = [ + dict(backend_args=None, type='LoadImageFromFile'), + dict(keep_ratio=True, scale=( + 800, + 800, + ), type='Resize'), + dict(type='LoadAnnotations', with_bbox=True), + dict( + meta_keys=( + 'img_id', + 'img_path', + 'ori_shape', + 'img_shape', + 'scale_factor', + ), + type='PackDetInputs'), +] +train_cfg = dict(max_epochs=12, type='EpochBasedTrainLoop', val_interval=1) +train_dataloader = dict( + batch_sampler=dict(type='AspectRatioBatchSampler'), + batch_size=4, + dataset=dict( + ann_file='annotations/instances_train2017.json', + backend_args=None, + data_prefix=dict(img='train2017/'), + data_root='data/coco/', + filter_cfg=dict(filter_empty_gt=True, min_size=32), + pipeline=[ + dict(backend_args=None, type='LoadImageFromFile'), + dict(type='LoadAnnotations', with_bbox=True), + dict(keep_ratio=True, scale=( + 1333, + 800, + ), type='Resize'), + dict(prob=0.5, type='RandomFlip'), + dict(type='PackDetInputs'), + ], + type='CocoDataset'), + num_workers=4, + persistent_workers=True, + sampler=dict(shuffle=True, type='DefaultSampler')) +train_pipeline = [ + dict(backend_args=None, type='LoadImageFromFile'), + dict(type='LoadAnnotations', with_bbox=True), + dict(keep_ratio=True, scale=( + 1333, + 800, + ), type='Resize'), + dict(prob=0.5, type='RandomFlip'), + dict(type='PackDetInputs'), +] +val_cfg = dict(type='ValLoop') +val_dataloader = dict( + batch_size=1, + dataset=dict( + ann_file='annotations/instances_val2017.json', + backend_args=None, + data_prefix=dict(img='val2017/'), + data_root='data/coco/', + pipeline=[ + dict(backend_args=None, type='LoadImageFromFile'), + dict(keep_ratio=True, scale=( + 800, + 800, + ), type='Resize'), + dict(type='LoadAnnotations', with_bbox=True), + dict( + meta_keys=( + 'img_id', + 'img_path', + 'ori_shape', + 'img_shape', + 'scale_factor', + ), + type='PackDetInputs'), + ], + test_mode=True, + type='CocoDataset'), + drop_last=False, + num_workers=2, + persistent_workers=True, + sampler=dict(shuffle=False, type='DefaultSampler')) +val_evaluator = dict( + ann_file='data/coco/annotations/instances_val2017.json', + backend_args=None, + format_only=False, + metric='bbox', + type='CocoMetric') +vis_backends = [ + dict(type='LocalVisBackend'), +] +visualizer = dict( + name='visualizer', + type='DetLocalVisualizer', + vis_backends=[ + dict(type='LocalVisBackend'), + ]) +work_dir = './' diff --git a/models/cv/detection/hrnet/ixrt/inference.py b/models/cv/detection/hrnet/ixrt/inference.py new file mode 100644 index 0000000000000000000000000000000000000000..f5e84d967d991eb4de2fb9eb111737d6b74f2976 --- /dev/null +++ b/models/cv/detection/hrnet/ixrt/inference.py @@ -0,0 +1,190 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import time +import argparse +import tensorrt +import torch +import torchvision +import numpy as np +from tensorrt import Dims +from cuda import cuda, cudart +from tqdm import tqdm +from mmdet.registry import RUNNERS +from mmengine.config import Config + +from common import create_engine_context, get_io_bindings + +def parse_args(): + parser = argparse.ArgumentParser() + + parser.add_argument("--engine", + type=str, + required=True, + help="igie engine path.") + + parser.add_argument("--batchsize", + type=int, + required=True, + help="inference batch size.") + + parser.add_argument("--datasets", + type=str, + required=True, + help="datasets path.") + + parser.add_argument("--input_name", + type=str, + required=True, + help="input name of the model.") + + parser.add_argument("--warmup", + type=int, + default=3, + help="number of warmup before test.") + + parser.add_argument("--acc_target", + type=float, + default=None, + help="Model inference Accuracy target.") + + parser.add_argument("--fps_target", + type=float, + default=None, + help="Model inference FPS target.") + + parser.add_argument("--perf_only", + type=bool, + default=False, + help="Run performance test only") + + args = parser.parse_args() + + return args + +def main(): + args = parse_args() + + batch_size = args.batchsize + + host_mem = tensorrt.IHostMemory + logger = tensorrt.Logger(tensorrt.Logger.ERROR) + + # Load Engine && I/O bindings + engine, context = create_engine_context(args.engine, logger) + inputs, outputs, allocations = get_io_bindings(engine) + + if args.warmup > 0: + print("\nWarm Start.") + for i in range(args.warmup): + context.execute_v2(allocations) + print("Warm Done.") + + # just run perf test + if args.perf_only: + torch.cuda.synchronize() + start_time = time.time() + + for i in range(10): + context.execute_v2(allocations) + + torch.cuda.synchronize() + end_time = time.time() + forward_time = end_time - start_time + num_samples = 10 * args.batchsize + fps = num_samples / forward_time + + print("FPS : ", fps) + print(f"Performance Check : Test {fps} >= target {args.fps_target}") + if fps >= args.fps_target: + print("pass!") + exit() + else: + print("failed!") + exit(1) + else: + # Runner config + cfg = Config.fromfile("fcos_hrnetv2p-w18-gn-head_4xb4-1x_coco.py") + cfg.work_dir = "./" + + cfg['test_dataloader']['batch_size'] = batch_size + cfg['test_dataloader']['dataset']['data_root'] = args.datasets + cfg['test_dataloader']['dataset']['data_prefix']['img'] = 'images/val2017/' + cfg['test_evaluator']['ann_file'] = os.path.join(args.datasets, 'annotations/instances_val2017.json') + cfg['log_level'] = 'ERROR' + + runner = RUNNERS.build(cfg) + + for input_data in tqdm(runner.test_dataloader): + + input_data = runner.model.data_preprocessor(input_data, False) + image = input_data['inputs'].cpu() + image = image.numpy().astype(inputs[0]["dtype"]) + pad_batch = len(image) != batch_size + if pad_batch: + origin_size = len(image) + image = np.resize(image, (batch_size, *image.shape[1:])) + image = np.ascontiguousarray(image) + + (err,) = cudart.cudaMemcpy( + inputs[0]["allocation"], + image, + image.nbytes, + cudart.cudaMemcpyKind.cudaMemcpyHostToDevice, + ) + assert err == cudart.cudaError_t.cudaSuccess + + context.execute_v2(allocations) + + cls_score = [] + box_reg = [] + score_factors = [] + for i in range(len(outputs)): + output = np.zeros(outputs[i]["shape"], outputs[i]["dtype"]) + (err,) = cudart.cudaMemcpy( + output, + outputs[i]["allocation"], + outputs[i]["nbytes"], + cudart.cudaMemcpyKind.cudaMemcpyDeviceToHost, + ) + assert err == cudart.cudaError_t.cudaSuccess + + if pad_batch: + output = output[:origin_size] + + output = torch.from_numpy(output) + + if output.shape[1] == 80: + cls_score.append(output) + elif output.shape[1] == 4: + box_reg.append(output) + else: + score_factors.append(output) + + batch_img_metas = [ + data_samples.metainfo for data_samples in input_data['data_samples'] + ] + + results_list = runner.model.bbox_head.predict_by_feat(cls_score, box_reg, score_factors, batch_img_metas=batch_img_metas, rescale=True) + + batch_data_samples = runner.model.add_pred_to_datasample(input_data['data_samples'], results_list) + + runner.test_evaluator.process(data_samples=batch_data_samples, data_batch=input_data) + + metrics = runner.test_evaluator.evaluate(len(runner.test_dataloader.dataset)) + +if __name__ == "__main__": + main() \ No newline at end of file diff --git a/models/cv/detection/hrnet/ixrt/requirements.txt b/models/cv/detection/hrnet/ixrt/requirements.txt new file mode 100644 index 0000000000000000000000000000000000000000..97ac9c0458744fb56d62781ffd96279f893817f3 --- /dev/null +++ b/models/cv/detection/hrnet/ixrt/requirements.txt @@ -0,0 +1,6 @@ +onnx +tqdm +onnxsim +mmdet +mmdeploy +mmengine diff --git a/models/cv/detection/hrnet/ixrt/scripts/infer_hrnet_fp16_accuracy.sh b/models/cv/detection/hrnet/ixrt/scripts/infer_hrnet_fp16_accuracy.sh new file mode 100644 index 0000000000000000000000000000000000000000..74f55c2970c1b3a2a8902b5b3884b2cecb4dfb3b --- /dev/null +++ b/models/cv/detection/hrnet/ixrt/scripts/infer_hrnet_fp16_accuracy.sh @@ -0,0 +1,34 @@ +#!/bin/bash + +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +batchsize=32 +model_path="hrnet_opt.onnx" +datasets_path=${DATASETS_DIR} + +# build engine +python3 build_engine.py \ + --model ${model_path} \ + --precision float16 \ + --engine hrnet.engine + + +# inference +python3 inference.py \ + --engine hrnet.engine \ + --batchsize ${batchsize} \ + --input_name input \ + --datasets ${datasets_path} \ No newline at end of file diff --git a/models/cv/detection/hrnet/ixrt/scripts/infer_hrnet_fp16_performance.sh b/models/cv/detection/hrnet/ixrt/scripts/infer_hrnet_fp16_performance.sh new file mode 100644 index 0000000000000000000000000000000000000000..ae6a5ef1529ff76de25c62be2b20eb003f79f678 --- /dev/null +++ b/models/cv/detection/hrnet/ixrt/scripts/infer_hrnet_fp16_performance.sh @@ -0,0 +1,36 @@ +#!/bin/bash + +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +batchsize=32 +model_path="hrnet_opt.onnx" +datasets_path=${DATASETS_DIR} + +# build engine +python3 build_engine.py \ + --model ${model_path} \ + --precision float16 \ + --engine hrnet.engine + + +# inference +python3 inference.py \ + --engine hrnet.engine \ + --batchsize ${batchsize} \ + --input_name input \ + --datasets ${datasets_path} \ + --perf_only True \ + --fps_target 70 \ No newline at end of file diff --git a/models/cv/detection/retinaface/ixrt/README.md b/models/cv/detection/retinaface/ixrt/README.md new file mode 100644 index 0000000000000000000000000000000000000000..8fff5439fd8e82e836e58abc7c38296f0ec27db5 --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/README.md @@ -0,0 +1,63 @@ +# RetinaFace + +## Description + +RetinaFace is an efficient single-stage face detection model that employs a multi-task learning strategy to simultaneously predict facial locations, landmarks, and 3D facial shapes. It utilizes feature pyramids and context modules to extract multi-scale features and employs a self-supervised mesh decoder to enhance detection accuracy. RetinaFace demonstrates excellent performance on datasets like WIDER FACE, supports real-time processing, and its code and datasets are publicly available for researchers. + +## Setup + +### Install + +```bash +# Install libGL +## CentOS +yum install -y mesa-libGL +## Ubuntu +apt install -y libgl1-mesa-dev + +pip3 install -r requirements.txt + +python3 setup.py build_ext --inplace +``` + +### Download + +Pretrained model: + +Dataset: to download the validation dataset. + +```bash +wget https://github.com/biubug6/Face-Detector-1MB-with-landmark/raw/master/weights/mobilenet0.25_Final.pth +``` + +### Model Conversion + +```bash +# export onnx model +python3 torch2onnx.py --model mobilenet0.25_Final.pth --onnx_model mnetv1_retinaface.onnx + +## Inference + +```bash +export DATASETS_DIR=/Path/to/widerface/ +export GT_DIR=../igie/ground_truth +``` + +### FP16 + +```bash +# Accuracy +bash scripts/infer_retinaface_fp16_accuracy.sh +# Performance +bash scripts/infer_retinaface_fp16_performance.sh +``` + +## Results + +| Model | BatchSize | Precision | FPS | Easy AP(%) | Medium AP (%) | Hard AP(%) | +| :--------: | :-------: | :-------: | :------: | :--------: | :-----------: | :--------: | +| RetinaFace | 32 | FP16 | 8536.367 | 80.84 | 69.34 | 37.31 | + +## Reference + +Face-Detector-1MB-with-landmark: diff --git a/models/cv/detection/retinaface/ixrt/__init__.py b/models/cv/detection/retinaface/ixrt/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/models/cv/detection/retinaface/ixrt/box_overlaps.c b/models/cv/detection/retinaface/ixrt/box_overlaps.c new file mode 100644 index 0000000000000000000000000000000000000000..364ceaae27f00b8616a3798d584cd799678913e5 --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/box_overlaps.c @@ -0,0 +1,7097 @@ +/* Generated by Cython 0.29.36 */ + +/* BEGIN: Cython Metadata +{ + "distutils": { + "depends": [ + "/usr/local/lib/python3.10/site-packages/numpy/core/include/numpy/arrayobject.h", + "/usr/local/lib/python3.10/site-packages/numpy/core/include/numpy/arrayscalars.h", + "/usr/local/lib/python3.10/site-packages/numpy/core/include/numpy/ndarrayobject.h", + "/usr/local/lib/python3.10/site-packages/numpy/core/include/numpy/ndarraytypes.h", + "/usr/local/lib/python3.10/site-packages/numpy/core/include/numpy/ufuncobject.h" + ], + "include_dirs": [ + "/usr/local/lib/python3.10/site-packages/numpy/core/include" + ], + "name": "bbox", + "sources": [ + "box_overlaps.pyx" + ] + }, + "module_name": "bbox" +} +END: Cython Metadata */ + +#ifndef PY_SSIZE_T_CLEAN +#define PY_SSIZE_T_CLEAN +#endif /* PY_SSIZE_T_CLEAN */ +#include "Python.h" +#ifndef Py_PYTHON_H + #error Python headers needed to compile C extensions, please install development version of Python. +#elif PY_VERSION_HEX < 0x02060000 || (0x03000000 <= PY_VERSION_HEX && PY_VERSION_HEX < 0x03030000) + #error Cython requires Python 2.6+ or Python 3.3+. +#else +#define CYTHON_ABI "0_29_36" +#define CYTHON_HEX_VERSION 0x001D24F0 +#define CYTHON_FUTURE_DIVISION 0 +#include +#ifndef offsetof + #define offsetof(type, member) ( (size_t) & ((type*)0) -> member ) +#endif +#if !defined(WIN32) && !defined(MS_WINDOWS) + #ifndef __stdcall + #define __stdcall + #endif + #ifndef __cdecl + #define __cdecl + #endif + #ifndef __fastcall + #define __fastcall + #endif +#endif +#ifndef DL_IMPORT + #define DL_IMPORT(t) t +#endif +#ifndef DL_EXPORT + #define DL_EXPORT(t) t +#endif +#define __PYX_COMMA , +#ifndef HAVE_LONG_LONG + #if PY_VERSION_HEX >= 0x02070000 + #define HAVE_LONG_LONG + #endif +#endif +#ifndef PY_LONG_LONG + #define PY_LONG_LONG LONG_LONG +#endif +#ifndef Py_HUGE_VAL + #define Py_HUGE_VAL HUGE_VAL +#endif +#ifdef PYPY_VERSION + #define CYTHON_COMPILING_IN_PYPY 1 + #define CYTHON_COMPILING_IN_PYSTON 0 + #define CYTHON_COMPILING_IN_CPYTHON 0 + #define CYTHON_COMPILING_IN_NOGIL 0 + #undef CYTHON_USE_TYPE_SLOTS + #define CYTHON_USE_TYPE_SLOTS 0 + #undef CYTHON_USE_PYTYPE_LOOKUP + #define CYTHON_USE_PYTYPE_LOOKUP 0 + #if PY_VERSION_HEX < 0x03050000 + #undef CYTHON_USE_ASYNC_SLOTS + #define CYTHON_USE_ASYNC_SLOTS 0 + #elif !defined(CYTHON_USE_ASYNC_SLOTS) + #define CYTHON_USE_ASYNC_SLOTS 1 + #endif + #undef CYTHON_USE_PYLIST_INTERNALS + #define CYTHON_USE_PYLIST_INTERNALS 0 + #undef CYTHON_USE_UNICODE_INTERNALS + #define CYTHON_USE_UNICODE_INTERNALS 0 + #undef CYTHON_USE_UNICODE_WRITER + #define CYTHON_USE_UNICODE_WRITER 0 + #undef CYTHON_USE_PYLONG_INTERNALS + #define CYTHON_USE_PYLONG_INTERNALS 0 + #undef CYTHON_AVOID_BORROWED_REFS + #define CYTHON_AVOID_BORROWED_REFS 1 + #undef CYTHON_ASSUME_SAFE_MACROS + #define CYTHON_ASSUME_SAFE_MACROS 0 + #undef CYTHON_UNPACK_METHODS + #define CYTHON_UNPACK_METHODS 0 + #undef CYTHON_FAST_THREAD_STATE + #define CYTHON_FAST_THREAD_STATE 0 + #undef CYTHON_FAST_PYCALL + #define CYTHON_FAST_PYCALL 0 + #if PY_VERSION_HEX < 0x03090000 + #undef CYTHON_PEP489_MULTI_PHASE_INIT + #define CYTHON_PEP489_MULTI_PHASE_INIT 0 + #elif !defined(CYTHON_PEP489_MULTI_PHASE_INIT) + #define CYTHON_PEP489_MULTI_PHASE_INIT 1 + #endif + #undef CYTHON_USE_TP_FINALIZE + #define CYTHON_USE_TP_FINALIZE (PY_VERSION_HEX >= 0x030400a1 && PYPY_VERSION_NUM >= 0x07030C00) + #undef CYTHON_USE_DICT_VERSIONS + #define CYTHON_USE_DICT_VERSIONS 0 + #undef CYTHON_USE_EXC_INFO_STACK + #define CYTHON_USE_EXC_INFO_STACK 0 + #ifndef CYTHON_UPDATE_DESCRIPTOR_DOC + #define CYTHON_UPDATE_DESCRIPTOR_DOC 0 + #endif +#elif defined(PYSTON_VERSION) + #define CYTHON_COMPILING_IN_PYPY 0 + #define CYTHON_COMPILING_IN_PYSTON 1 + #define CYTHON_COMPILING_IN_CPYTHON 0 + #define CYTHON_COMPILING_IN_NOGIL 0 + #ifndef CYTHON_USE_TYPE_SLOTS + #define CYTHON_USE_TYPE_SLOTS 1 + #endif + #undef CYTHON_USE_PYTYPE_LOOKUP + #define CYTHON_USE_PYTYPE_LOOKUP 0 + #undef CYTHON_USE_ASYNC_SLOTS + #define CYTHON_USE_ASYNC_SLOTS 0 + #undef CYTHON_USE_PYLIST_INTERNALS + #define CYTHON_USE_PYLIST_INTERNALS 0 + #ifndef CYTHON_USE_UNICODE_INTERNALS + #define CYTHON_USE_UNICODE_INTERNALS 1 + #endif + #undef CYTHON_USE_UNICODE_WRITER + #define CYTHON_USE_UNICODE_WRITER 0 + #undef CYTHON_USE_PYLONG_INTERNALS + #define CYTHON_USE_PYLONG_INTERNALS 0 + #ifndef CYTHON_AVOID_BORROWED_REFS + #define CYTHON_AVOID_BORROWED_REFS 0 + #endif + #ifndef CYTHON_ASSUME_SAFE_MACROS + #define CYTHON_ASSUME_SAFE_MACROS 1 + #endif + #ifndef CYTHON_UNPACK_METHODS + #define CYTHON_UNPACK_METHODS 1 + #endif + #undef CYTHON_FAST_THREAD_STATE + #define CYTHON_FAST_THREAD_STATE 0 + #undef CYTHON_FAST_PYCALL + #define CYTHON_FAST_PYCALL 0 + #undef CYTHON_PEP489_MULTI_PHASE_INIT + #define CYTHON_PEP489_MULTI_PHASE_INIT 0 + #undef CYTHON_USE_TP_FINALIZE + #define CYTHON_USE_TP_FINALIZE 0 + #undef CYTHON_USE_DICT_VERSIONS + #define CYTHON_USE_DICT_VERSIONS 0 + #undef CYTHON_USE_EXC_INFO_STACK + #define CYTHON_USE_EXC_INFO_STACK 0 + #ifndef CYTHON_UPDATE_DESCRIPTOR_DOC + #define CYTHON_UPDATE_DESCRIPTOR_DOC 0 + #endif +#elif defined(PY_NOGIL) + #define CYTHON_COMPILING_IN_PYPY 0 + #define CYTHON_COMPILING_IN_PYSTON 0 + #define CYTHON_COMPILING_IN_CPYTHON 0 + #define CYTHON_COMPILING_IN_NOGIL 1 + #ifndef CYTHON_USE_TYPE_SLOTS + #define CYTHON_USE_TYPE_SLOTS 1 + #endif + #undef CYTHON_USE_PYTYPE_LOOKUP + #define CYTHON_USE_PYTYPE_LOOKUP 0 + #ifndef CYTHON_USE_ASYNC_SLOTS + #define CYTHON_USE_ASYNC_SLOTS 1 + #endif + #undef CYTHON_USE_PYLIST_INTERNALS + #define CYTHON_USE_PYLIST_INTERNALS 0 + #ifndef CYTHON_USE_UNICODE_INTERNALS + #define CYTHON_USE_UNICODE_INTERNALS 1 + #endif + #undef CYTHON_USE_UNICODE_WRITER + #define CYTHON_USE_UNICODE_WRITER 0 + #undef CYTHON_USE_PYLONG_INTERNALS + #define CYTHON_USE_PYLONG_INTERNALS 0 + #ifndef CYTHON_AVOID_BORROWED_REFS + #define CYTHON_AVOID_BORROWED_REFS 0 + #endif + #ifndef CYTHON_ASSUME_SAFE_MACROS + #define CYTHON_ASSUME_SAFE_MACROS 1 + #endif + #ifndef CYTHON_UNPACK_METHODS + #define CYTHON_UNPACK_METHODS 1 + #endif + #undef CYTHON_FAST_THREAD_STATE + #define CYTHON_FAST_THREAD_STATE 0 + #undef CYTHON_FAST_PYCALL + #define CYTHON_FAST_PYCALL 0 + #ifndef CYTHON_PEP489_MULTI_PHASE_INIT + #define CYTHON_PEP489_MULTI_PHASE_INIT 1 + #endif + #ifndef CYTHON_USE_TP_FINALIZE + #define CYTHON_USE_TP_FINALIZE 1 + #endif + #undef CYTHON_USE_DICT_VERSIONS + #define CYTHON_USE_DICT_VERSIONS 0 + #undef CYTHON_USE_EXC_INFO_STACK + #define CYTHON_USE_EXC_INFO_STACK 0 +#else + #define CYTHON_COMPILING_IN_PYPY 0 + #define CYTHON_COMPILING_IN_PYSTON 0 + #define CYTHON_COMPILING_IN_CPYTHON 1 + #define CYTHON_COMPILING_IN_NOGIL 0 + #ifndef CYTHON_USE_TYPE_SLOTS + #define CYTHON_USE_TYPE_SLOTS 1 + #endif + #if PY_VERSION_HEX < 0x02070000 + #undef CYTHON_USE_PYTYPE_LOOKUP + #define CYTHON_USE_PYTYPE_LOOKUP 0 + #elif !defined(CYTHON_USE_PYTYPE_LOOKUP) + #define CYTHON_USE_PYTYPE_LOOKUP 1 + #endif + #if PY_MAJOR_VERSION < 3 + #undef CYTHON_USE_ASYNC_SLOTS + #define CYTHON_USE_ASYNC_SLOTS 0 + #elif !defined(CYTHON_USE_ASYNC_SLOTS) + #define CYTHON_USE_ASYNC_SLOTS 1 + #endif + #if PY_VERSION_HEX < 0x02070000 + #undef CYTHON_USE_PYLONG_INTERNALS + #define CYTHON_USE_PYLONG_INTERNALS 0 + #elif !defined(CYTHON_USE_PYLONG_INTERNALS) + #define CYTHON_USE_PYLONG_INTERNALS (PY_VERSION_HEX < 0x030C00A5) + #endif + #ifndef CYTHON_USE_PYLIST_INTERNALS + #define CYTHON_USE_PYLIST_INTERNALS 1 + #endif + #ifndef CYTHON_USE_UNICODE_INTERNALS + #define CYTHON_USE_UNICODE_INTERNALS 1 + #endif + #if PY_VERSION_HEX < 0x030300F0 || PY_VERSION_HEX >= 0x030B00A2 + #undef CYTHON_USE_UNICODE_WRITER + #define CYTHON_USE_UNICODE_WRITER 0 + #elif !defined(CYTHON_USE_UNICODE_WRITER) + #define CYTHON_USE_UNICODE_WRITER 1 + #endif + #ifndef CYTHON_AVOID_BORROWED_REFS + #define CYTHON_AVOID_BORROWED_REFS 0 + #endif + #ifndef CYTHON_ASSUME_SAFE_MACROS + #define CYTHON_ASSUME_SAFE_MACROS 1 + #endif + #ifndef CYTHON_UNPACK_METHODS + #define CYTHON_UNPACK_METHODS 1 + #endif + #if PY_VERSION_HEX >= 0x030B00A4 + #undef CYTHON_FAST_THREAD_STATE + #define CYTHON_FAST_THREAD_STATE 0 + #elif !defined(CYTHON_FAST_THREAD_STATE) + #define CYTHON_FAST_THREAD_STATE 1 + #endif + #ifndef CYTHON_FAST_PYCALL + #define CYTHON_FAST_PYCALL (PY_VERSION_HEX < 0x030A0000) + #endif + #ifndef CYTHON_PEP489_MULTI_PHASE_INIT + #define CYTHON_PEP489_MULTI_PHASE_INIT (PY_VERSION_HEX >= 0x03050000) + #endif + #ifndef CYTHON_USE_TP_FINALIZE + #define CYTHON_USE_TP_FINALIZE (PY_VERSION_HEX >= 0x030400a1) + #endif + #ifndef CYTHON_USE_DICT_VERSIONS + #define CYTHON_USE_DICT_VERSIONS ((PY_VERSION_HEX >= 0x030600B1) && (PY_VERSION_HEX < 0x030C00A5)) + #endif + #if PY_VERSION_HEX >= 0x030B00A4 + #undef CYTHON_USE_EXC_INFO_STACK + #define CYTHON_USE_EXC_INFO_STACK 0 + #elif !defined(CYTHON_USE_EXC_INFO_STACK) + #define CYTHON_USE_EXC_INFO_STACK (PY_VERSION_HEX >= 0x030700A3) + #endif + #ifndef CYTHON_UPDATE_DESCRIPTOR_DOC + #define CYTHON_UPDATE_DESCRIPTOR_DOC 1 + #endif +#endif +#if !defined(CYTHON_FAST_PYCCALL) +#define CYTHON_FAST_PYCCALL (CYTHON_FAST_PYCALL && PY_VERSION_HEX >= 0x030600B1) +#endif +#if CYTHON_USE_PYLONG_INTERNALS + #if PY_MAJOR_VERSION < 3 + #include "longintrepr.h" + #endif + #undef SHIFT + #undef BASE + #undef MASK + #ifdef SIZEOF_VOID_P + enum { __pyx_check_sizeof_voidp = 1 / (int)(SIZEOF_VOID_P == sizeof(void*)) }; + #endif +#endif +#ifndef __has_attribute + #define __has_attribute(x) 0 +#endif +#ifndef __has_cpp_attribute + #define __has_cpp_attribute(x) 0 +#endif +#ifndef CYTHON_RESTRICT + #if defined(__GNUC__) + #define CYTHON_RESTRICT __restrict__ + #elif defined(_MSC_VER) && _MSC_VER >= 1400 + #define CYTHON_RESTRICT __restrict + #elif defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L + #define CYTHON_RESTRICT restrict + #else + #define CYTHON_RESTRICT + #endif +#endif +#ifndef CYTHON_UNUSED +# if defined(__GNUC__) +# if !(defined(__cplusplus)) || (__GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ >= 4)) +# define CYTHON_UNUSED __attribute__ ((__unused__)) +# else +# define CYTHON_UNUSED +# endif +# elif defined(__ICC) || (defined(__INTEL_COMPILER) && !defined(_MSC_VER)) +# define CYTHON_UNUSED __attribute__ ((__unused__)) +# else +# define CYTHON_UNUSED +# endif +#endif +#ifndef CYTHON_MAYBE_UNUSED_VAR +# if defined(__cplusplus) + template void CYTHON_MAYBE_UNUSED_VAR( const T& ) { } +# else +# define CYTHON_MAYBE_UNUSED_VAR(x) (void)(x) +# endif +#endif +#ifndef CYTHON_NCP_UNUSED +# if CYTHON_COMPILING_IN_CPYTHON +# define CYTHON_NCP_UNUSED +# else +# define CYTHON_NCP_UNUSED CYTHON_UNUSED +# endif +#endif +#define __Pyx_void_to_None(void_result) ((void)(void_result), Py_INCREF(Py_None), Py_None) +#ifdef _MSC_VER + #ifndef _MSC_STDINT_H_ + #if _MSC_VER < 1300 + typedef unsigned char uint8_t; + typedef unsigned int uint32_t; + #else + typedef unsigned __int8 uint8_t; + typedef unsigned __int32 uint32_t; + #endif + #endif +#else + #include +#endif +#ifndef CYTHON_FALLTHROUGH + #if defined(__cplusplus) && __cplusplus >= 201103L + #if __has_cpp_attribute(fallthrough) + #define CYTHON_FALLTHROUGH [[fallthrough]] + #elif __has_cpp_attribute(clang::fallthrough) + #define CYTHON_FALLTHROUGH [[clang::fallthrough]] + #elif __has_cpp_attribute(gnu::fallthrough) + #define CYTHON_FALLTHROUGH [[gnu::fallthrough]] + #endif + #endif + #ifndef CYTHON_FALLTHROUGH + #if __has_attribute(fallthrough) + #define CYTHON_FALLTHROUGH __attribute__((fallthrough)) + #else + #define CYTHON_FALLTHROUGH + #endif + #endif + #if defined(__clang__ ) && defined(__apple_build_version__) + #if __apple_build_version__ < 7000000 + #undef CYTHON_FALLTHROUGH + #define CYTHON_FALLTHROUGH + #endif + #endif +#endif + +#ifndef CYTHON_INLINE + #if defined(__clang__) + #define CYTHON_INLINE __inline__ __attribute__ ((__unused__)) + #elif defined(__GNUC__) + #define CYTHON_INLINE __inline__ + #elif defined(_MSC_VER) + #define CYTHON_INLINE __inline + #elif defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L + #define CYTHON_INLINE inline + #else + #define CYTHON_INLINE + #endif +#endif + +#define __PYX_BUILD_PY_SSIZE_T "n" +#define CYTHON_FORMAT_SSIZE_T "z" +#if PY_MAJOR_VERSION < 3 + #define __Pyx_BUILTIN_MODULE_NAME "__builtin__" + #define __Pyx_PyCode_New(a, k, l, s, f, code, c, n, v, fv, cell, fn, name, fline, lnos)\ + PyCode_New(a+k, l, s, f, code, c, n, v, fv, cell, fn, name, fline, lnos) + #define __Pyx_DefaultClassType PyClass_Type +#else + #define __Pyx_BUILTIN_MODULE_NAME "builtins" + #define __Pyx_DefaultClassType PyType_Type +#if PY_VERSION_HEX >= 0x030B00A1 + static CYTHON_INLINE PyCodeObject* __Pyx_PyCode_New(int a, int k, int l, int s, int f, + PyObject *code, PyObject *c, PyObject* n, PyObject *v, + PyObject *fv, PyObject *cell, PyObject* fn, + PyObject *name, int fline, PyObject *lnos) { + PyObject *kwds=NULL, *argcount=NULL, *posonlyargcount=NULL, *kwonlyargcount=NULL; + PyObject *nlocals=NULL, *stacksize=NULL, *flags=NULL, *replace=NULL, *call_result=NULL, *empty=NULL; + const char *fn_cstr=NULL; + const char *name_cstr=NULL; + PyCodeObject* co=NULL; + PyObject *type, *value, *traceback; + PyErr_Fetch(&type, &value, &traceback); + if (!(kwds=PyDict_New())) goto end; + if (!(argcount=PyLong_FromLong(a))) goto end; + if (PyDict_SetItemString(kwds, "co_argcount", argcount) != 0) goto end; + if (!(posonlyargcount=PyLong_FromLong(0))) goto end; + if (PyDict_SetItemString(kwds, "co_posonlyargcount", posonlyargcount) != 0) goto end; + if (!(kwonlyargcount=PyLong_FromLong(k))) goto end; + if (PyDict_SetItemString(kwds, "co_kwonlyargcount", kwonlyargcount) != 0) goto end; + if (!(nlocals=PyLong_FromLong(l))) goto end; + if (PyDict_SetItemString(kwds, "co_nlocals", nlocals) != 0) goto end; + if (!(stacksize=PyLong_FromLong(s))) goto end; + if (PyDict_SetItemString(kwds, "co_stacksize", stacksize) != 0) goto end; + if (!(flags=PyLong_FromLong(f))) goto end; + if (PyDict_SetItemString(kwds, "co_flags", flags) != 0) goto end; + if (PyDict_SetItemString(kwds, "co_code", code) != 0) goto end; + if (PyDict_SetItemString(kwds, "co_consts", c) != 0) goto end; + if (PyDict_SetItemString(kwds, "co_names", n) != 0) goto end; + if (PyDict_SetItemString(kwds, "co_varnames", v) != 0) goto end; + if (PyDict_SetItemString(kwds, "co_freevars", fv) != 0) goto end; + if (PyDict_SetItemString(kwds, "co_cellvars", cell) != 0) goto end; + if (PyDict_SetItemString(kwds, "co_linetable", lnos) != 0) goto end; + if (!(fn_cstr=PyUnicode_AsUTF8AndSize(fn, NULL))) goto end; + if (!(name_cstr=PyUnicode_AsUTF8AndSize(name, NULL))) goto end; + if (!(co = PyCode_NewEmpty(fn_cstr, name_cstr, fline))) goto end; + if (!(replace = PyObject_GetAttrString((PyObject*)co, "replace"))) goto cleanup_code_too; + if (!(empty = PyTuple_New(0))) goto cleanup_code_too; // unfortunately __pyx_empty_tuple isn't available here + if (!(call_result = PyObject_Call(replace, empty, kwds))) goto cleanup_code_too; + Py_XDECREF((PyObject*)co); + co = (PyCodeObject*)call_result; + call_result = NULL; + if (0) { + cleanup_code_too: + Py_XDECREF((PyObject*)co); + co = NULL; + } + end: + Py_XDECREF(kwds); + Py_XDECREF(argcount); + Py_XDECREF(posonlyargcount); + Py_XDECREF(kwonlyargcount); + Py_XDECREF(nlocals); + Py_XDECREF(stacksize); + Py_XDECREF(replace); + Py_XDECREF(call_result); + Py_XDECREF(empty); + if (type) { + PyErr_Restore(type, value, traceback); + } + return co; + } +#else + #define __Pyx_PyCode_New(a, k, l, s, f, code, c, n, v, fv, cell, fn, name, fline, lnos)\ + PyCode_New(a, k, l, s, f, code, c, n, v, fv, cell, fn, name, fline, lnos) +#endif + #define __Pyx_DefaultClassType PyType_Type +#endif +#if PY_VERSION_HEX >= 0x030900F0 && !CYTHON_COMPILING_IN_PYPY + #define __Pyx_PyObject_GC_IsFinalized(o) PyObject_GC_IsFinalized(o) +#else + #define __Pyx_PyObject_GC_IsFinalized(o) _PyGC_FINALIZED(o) +#endif +#ifndef Py_TPFLAGS_CHECKTYPES + #define Py_TPFLAGS_CHECKTYPES 0 +#endif +#ifndef Py_TPFLAGS_HAVE_INDEX + #define Py_TPFLAGS_HAVE_INDEX 0 +#endif +#ifndef Py_TPFLAGS_HAVE_NEWBUFFER + #define Py_TPFLAGS_HAVE_NEWBUFFER 0 +#endif +#ifndef Py_TPFLAGS_HAVE_FINALIZE + #define Py_TPFLAGS_HAVE_FINALIZE 0 +#endif +#ifndef METH_STACKLESS + #define METH_STACKLESS 0 +#endif +#if PY_VERSION_HEX <= 0x030700A3 || !defined(METH_FASTCALL) + #ifndef METH_FASTCALL + #define METH_FASTCALL 0x80 + #endif + typedef PyObject *(*__Pyx_PyCFunctionFast) (PyObject *self, PyObject *const *args, Py_ssize_t nargs); + typedef PyObject *(*__Pyx_PyCFunctionFastWithKeywords) (PyObject *self, PyObject *const *args, + Py_ssize_t nargs, PyObject *kwnames); +#else + #define __Pyx_PyCFunctionFast _PyCFunctionFast + #define __Pyx_PyCFunctionFastWithKeywords _PyCFunctionFastWithKeywords +#endif +#if CYTHON_FAST_PYCCALL +#define __Pyx_PyFastCFunction_Check(func)\ + ((PyCFunction_Check(func) && (METH_FASTCALL == (PyCFunction_GET_FLAGS(func) & ~(METH_CLASS | METH_STATIC | METH_COEXIST | METH_KEYWORDS | METH_STACKLESS))))) +#else +#define __Pyx_PyFastCFunction_Check(func) 0 +#endif +#if CYTHON_COMPILING_IN_PYPY && !defined(PyObject_Malloc) + #define PyObject_Malloc(s) PyMem_Malloc(s) + #define PyObject_Free(p) PyMem_Free(p) + #define PyObject_Realloc(p) PyMem_Realloc(p) +#endif +#if CYTHON_COMPILING_IN_CPYTHON && PY_VERSION_HEX < 0x030400A1 + #define PyMem_RawMalloc(n) PyMem_Malloc(n) + #define PyMem_RawRealloc(p, n) PyMem_Realloc(p, n) + #define PyMem_RawFree(p) PyMem_Free(p) +#endif +#if CYTHON_COMPILING_IN_PYSTON + #define __Pyx_PyCode_HasFreeVars(co) PyCode_HasFreeVars(co) + #define __Pyx_PyFrame_SetLineNumber(frame, lineno) PyFrame_SetLineNumber(frame, lineno) +#else + #define __Pyx_PyCode_HasFreeVars(co) (PyCode_GetNumFree(co) > 0) + #define __Pyx_PyFrame_SetLineNumber(frame, lineno) (frame)->f_lineno = (lineno) +#endif +#if !CYTHON_FAST_THREAD_STATE || PY_VERSION_HEX < 0x02070000 + #define __Pyx_PyThreadState_Current PyThreadState_GET() +#elif PY_VERSION_HEX >= 0x03060000 + #define __Pyx_PyThreadState_Current _PyThreadState_UncheckedGet() +#elif PY_VERSION_HEX >= 0x03000000 + #define __Pyx_PyThreadState_Current PyThreadState_GET() +#else + #define __Pyx_PyThreadState_Current _PyThreadState_Current +#endif +#if PY_VERSION_HEX < 0x030700A2 && !defined(PyThread_tss_create) && !defined(Py_tss_NEEDS_INIT) +#include "pythread.h" +#define Py_tss_NEEDS_INIT 0 +typedef int Py_tss_t; +static CYTHON_INLINE int PyThread_tss_create(Py_tss_t *key) { + *key = PyThread_create_key(); + return 0; +} +static CYTHON_INLINE Py_tss_t * PyThread_tss_alloc(void) { + Py_tss_t *key = (Py_tss_t *)PyObject_Malloc(sizeof(Py_tss_t)); + *key = Py_tss_NEEDS_INIT; + return key; +} +static CYTHON_INLINE void PyThread_tss_free(Py_tss_t *key) { + PyObject_Free(key); +} +static CYTHON_INLINE int PyThread_tss_is_created(Py_tss_t *key) { + return *key != Py_tss_NEEDS_INIT; +} +static CYTHON_INLINE void PyThread_tss_delete(Py_tss_t *key) { + PyThread_delete_key(*key); + *key = Py_tss_NEEDS_INIT; +} +static CYTHON_INLINE int PyThread_tss_set(Py_tss_t *key, void *value) { + return PyThread_set_key_value(*key, value); +} +static CYTHON_INLINE void * PyThread_tss_get(Py_tss_t *key) { + return PyThread_get_key_value(*key); +} +#endif +#if CYTHON_COMPILING_IN_CPYTHON || defined(_PyDict_NewPresized) +#define __Pyx_PyDict_NewPresized(n) ((n <= 8) ? PyDict_New() : _PyDict_NewPresized(n)) +#else +#define __Pyx_PyDict_NewPresized(n) PyDict_New() +#endif +#if PY_MAJOR_VERSION >= 3 || CYTHON_FUTURE_DIVISION + #define __Pyx_PyNumber_Divide(x,y) PyNumber_TrueDivide(x,y) + #define __Pyx_PyNumber_InPlaceDivide(x,y) PyNumber_InPlaceTrueDivide(x,y) +#else + #define __Pyx_PyNumber_Divide(x,y) PyNumber_Divide(x,y) + #define __Pyx_PyNumber_InPlaceDivide(x,y) PyNumber_InPlaceDivide(x,y) +#endif +#if CYTHON_COMPILING_IN_CPYTHON && PY_VERSION_HEX >= 0x030500A1 && CYTHON_USE_UNICODE_INTERNALS +#define __Pyx_PyDict_GetItemStr(dict, name) _PyDict_GetItem_KnownHash(dict, name, ((PyASCIIObject *) name)->hash) +#else +#define __Pyx_PyDict_GetItemStr(dict, name) PyDict_GetItem(dict, name) +#endif +#if PY_VERSION_HEX > 0x03030000 && defined(PyUnicode_KIND) + #define CYTHON_PEP393_ENABLED 1 + #if PY_VERSION_HEX >= 0x030C0000 + #define __Pyx_PyUnicode_READY(op) (0) + #else + #define __Pyx_PyUnicode_READY(op) (likely(PyUnicode_IS_READY(op)) ?\ + 0 : _PyUnicode_Ready((PyObject *)(op))) + #endif + #define __Pyx_PyUnicode_GET_LENGTH(u) PyUnicode_GET_LENGTH(u) + #define __Pyx_PyUnicode_READ_CHAR(u, i) PyUnicode_READ_CHAR(u, i) + #define __Pyx_PyUnicode_MAX_CHAR_VALUE(u) PyUnicode_MAX_CHAR_VALUE(u) + #define __Pyx_PyUnicode_KIND(u) PyUnicode_KIND(u) + #define __Pyx_PyUnicode_DATA(u) PyUnicode_DATA(u) + #define __Pyx_PyUnicode_READ(k, d, i) PyUnicode_READ(k, d, i) + #define __Pyx_PyUnicode_WRITE(k, d, i, ch) PyUnicode_WRITE(k, d, i, ch) + #if PY_VERSION_HEX >= 0x030C0000 + #define __Pyx_PyUnicode_IS_TRUE(u) (0 != PyUnicode_GET_LENGTH(u)) + #else + #if CYTHON_COMPILING_IN_CPYTHON && PY_VERSION_HEX >= 0x03090000 + #define __Pyx_PyUnicode_IS_TRUE(u) (0 != (likely(PyUnicode_IS_READY(u)) ? PyUnicode_GET_LENGTH(u) : ((PyCompactUnicodeObject *)(u))->wstr_length)) + #else + #define __Pyx_PyUnicode_IS_TRUE(u) (0 != (likely(PyUnicode_IS_READY(u)) ? PyUnicode_GET_LENGTH(u) : PyUnicode_GET_SIZE(u))) + #endif + #endif +#else + #define CYTHON_PEP393_ENABLED 0 + #define PyUnicode_1BYTE_KIND 1 + #define PyUnicode_2BYTE_KIND 2 + #define PyUnicode_4BYTE_KIND 4 + #define __Pyx_PyUnicode_READY(op) (0) + #define __Pyx_PyUnicode_GET_LENGTH(u) PyUnicode_GET_SIZE(u) + #define __Pyx_PyUnicode_READ_CHAR(u, i) ((Py_UCS4)(PyUnicode_AS_UNICODE(u)[i])) + #define __Pyx_PyUnicode_MAX_CHAR_VALUE(u) ((sizeof(Py_UNICODE) == 2) ? 65535 : 1114111) + #define __Pyx_PyUnicode_KIND(u) (sizeof(Py_UNICODE)) + #define __Pyx_PyUnicode_DATA(u) ((void*)PyUnicode_AS_UNICODE(u)) + #define __Pyx_PyUnicode_READ(k, d, i) ((void)(k), (Py_UCS4)(((Py_UNICODE*)d)[i])) + #define __Pyx_PyUnicode_WRITE(k, d, i, ch) (((void)(k)), ((Py_UNICODE*)d)[i] = ch) + #define __Pyx_PyUnicode_IS_TRUE(u) (0 != PyUnicode_GET_SIZE(u)) +#endif +#if CYTHON_COMPILING_IN_PYPY + #define __Pyx_PyUnicode_Concat(a, b) PyNumber_Add(a, b) + #define __Pyx_PyUnicode_ConcatSafe(a, b) PyNumber_Add(a, b) +#else + #define __Pyx_PyUnicode_Concat(a, b) PyUnicode_Concat(a, b) + #define __Pyx_PyUnicode_ConcatSafe(a, b) ((unlikely((a) == Py_None) || unlikely((b) == Py_None)) ?\ + PyNumber_Add(a, b) : __Pyx_PyUnicode_Concat(a, b)) +#endif +#if CYTHON_COMPILING_IN_PYPY && !defined(PyUnicode_Contains) + #define PyUnicode_Contains(u, s) PySequence_Contains(u, s) +#endif +#if CYTHON_COMPILING_IN_PYPY && !defined(PyByteArray_Check) + #define PyByteArray_Check(obj) PyObject_TypeCheck(obj, &PyByteArray_Type) +#endif +#if CYTHON_COMPILING_IN_PYPY && !defined(PyObject_Format) + #define PyObject_Format(obj, fmt) PyObject_CallMethod(obj, "__format__", "O", fmt) +#endif +#define __Pyx_PyString_FormatSafe(a, b) ((unlikely((a) == Py_None || (PyString_Check(b) && !PyString_CheckExact(b)))) ? PyNumber_Remainder(a, b) : __Pyx_PyString_Format(a, b)) +#define __Pyx_PyUnicode_FormatSafe(a, b) ((unlikely((a) == Py_None || (PyUnicode_Check(b) && !PyUnicode_CheckExact(b)))) ? PyNumber_Remainder(a, b) : PyUnicode_Format(a, b)) +#if PY_MAJOR_VERSION >= 3 + #define __Pyx_PyString_Format(a, b) PyUnicode_Format(a, b) +#else + #define __Pyx_PyString_Format(a, b) PyString_Format(a, b) +#endif +#if PY_MAJOR_VERSION < 3 && !defined(PyObject_ASCII) + #define PyObject_ASCII(o) PyObject_Repr(o) +#endif +#if PY_MAJOR_VERSION >= 3 + #define PyBaseString_Type PyUnicode_Type + #define PyStringObject PyUnicodeObject + #define PyString_Type PyUnicode_Type + #define PyString_Check PyUnicode_Check + #define PyString_CheckExact PyUnicode_CheckExact +#ifndef PyObject_Unicode + #define PyObject_Unicode PyObject_Str +#endif +#endif +#if PY_MAJOR_VERSION >= 3 + #define __Pyx_PyBaseString_Check(obj) PyUnicode_Check(obj) + #define __Pyx_PyBaseString_CheckExact(obj) PyUnicode_CheckExact(obj) +#else + #define __Pyx_PyBaseString_Check(obj) (PyString_Check(obj) || PyUnicode_Check(obj)) + #define __Pyx_PyBaseString_CheckExact(obj) (PyString_CheckExact(obj) || PyUnicode_CheckExact(obj)) +#endif +#ifndef PySet_CheckExact + #define PySet_CheckExact(obj) (Py_TYPE(obj) == &PySet_Type) +#endif +#if PY_VERSION_HEX >= 0x030900A4 + #define __Pyx_SET_REFCNT(obj, refcnt) Py_SET_REFCNT(obj, refcnt) + #define __Pyx_SET_SIZE(obj, size) Py_SET_SIZE(obj, size) +#else + #define __Pyx_SET_REFCNT(obj, refcnt) Py_REFCNT(obj) = (refcnt) + #define __Pyx_SET_SIZE(obj, size) Py_SIZE(obj) = (size) +#endif +#if CYTHON_ASSUME_SAFE_MACROS + #define __Pyx_PySequence_SIZE(seq) Py_SIZE(seq) +#else + #define __Pyx_PySequence_SIZE(seq) PySequence_Size(seq) +#endif +#if PY_MAJOR_VERSION >= 3 + #define PyIntObject PyLongObject + #define PyInt_Type PyLong_Type + #define PyInt_Check(op) PyLong_Check(op) + #define PyInt_CheckExact(op) PyLong_CheckExact(op) + #define PyInt_FromString PyLong_FromString + #define PyInt_FromUnicode PyLong_FromUnicode + #define PyInt_FromLong PyLong_FromLong + #define PyInt_FromSize_t PyLong_FromSize_t + #define PyInt_FromSsize_t PyLong_FromSsize_t + #define PyInt_AsLong PyLong_AsLong + #define PyInt_AS_LONG PyLong_AS_LONG + #define PyInt_AsSsize_t PyLong_AsSsize_t + #define PyInt_AsUnsignedLongMask PyLong_AsUnsignedLongMask + #define PyInt_AsUnsignedLongLongMask PyLong_AsUnsignedLongLongMask + #define PyNumber_Int PyNumber_Long +#endif +#if PY_MAJOR_VERSION >= 3 + #define PyBoolObject PyLongObject +#endif +#if PY_MAJOR_VERSION >= 3 && CYTHON_COMPILING_IN_PYPY + #ifndef PyUnicode_InternFromString + #define PyUnicode_InternFromString(s) PyUnicode_FromString(s) + #endif +#endif +#if PY_VERSION_HEX < 0x030200A4 + typedef long Py_hash_t; + #define __Pyx_PyInt_FromHash_t PyInt_FromLong + #define __Pyx_PyInt_AsHash_t __Pyx_PyIndex_AsHash_t +#else + #define __Pyx_PyInt_FromHash_t PyInt_FromSsize_t + #define __Pyx_PyInt_AsHash_t __Pyx_PyIndex_AsSsize_t +#endif +#if PY_MAJOR_VERSION >= 3 + #define __Pyx_PyMethod_New(func, self, klass) ((self) ? ((void)(klass), PyMethod_New(func, self)) : __Pyx_NewRef(func)) +#else + #define __Pyx_PyMethod_New(func, self, klass) PyMethod_New(func, self, klass) +#endif +#if CYTHON_USE_ASYNC_SLOTS + #if PY_VERSION_HEX >= 0x030500B1 + #define __Pyx_PyAsyncMethodsStruct PyAsyncMethods + #define __Pyx_PyType_AsAsync(obj) (Py_TYPE(obj)->tp_as_async) + #else + #define __Pyx_PyType_AsAsync(obj) ((__Pyx_PyAsyncMethodsStruct*) (Py_TYPE(obj)->tp_reserved)) + #endif +#else + #define __Pyx_PyType_AsAsync(obj) NULL +#endif +#ifndef __Pyx_PyAsyncMethodsStruct + typedef struct { + unaryfunc am_await; + unaryfunc am_aiter; + unaryfunc am_anext; + } __Pyx_PyAsyncMethodsStruct; +#endif + +#if defined(_WIN32) || defined(WIN32) || defined(MS_WINDOWS) + #if !defined(_USE_MATH_DEFINES) + #define _USE_MATH_DEFINES + #endif +#endif +#include +#ifdef NAN +#define __PYX_NAN() ((float) NAN) +#else +static CYTHON_INLINE float __PYX_NAN() { + float value; + memset(&value, 0xFF, sizeof(value)); + return value; +} +#endif +#if defined(__CYGWIN__) && defined(_LDBL_EQ_DBL) +#define __Pyx_truncl trunc +#else +#define __Pyx_truncl truncl +#endif + +#define __PYX_MARK_ERR_POS(f_index, lineno) \ + { __pyx_filename = __pyx_f[f_index]; (void)__pyx_filename; __pyx_lineno = lineno; (void)__pyx_lineno; __pyx_clineno = __LINE__; (void)__pyx_clineno; } +#define __PYX_ERR(f_index, lineno, Ln_error) \ + { __PYX_MARK_ERR_POS(f_index, lineno) goto Ln_error; } + +#ifndef __PYX_EXTERN_C + #ifdef __cplusplus + #define __PYX_EXTERN_C extern "C" + #else + #define __PYX_EXTERN_C extern + #endif +#endif + +#define __PYX_HAVE__bbox +#define __PYX_HAVE_API__bbox +/* Early includes */ +#include +#include +#include "numpy/arrayobject.h" +#include "numpy/ndarrayobject.h" +#include "numpy/ndarraytypes.h" +#include "numpy/arrayscalars.h" +#include "numpy/ufuncobject.h" + + /* NumPy API declarations from "numpy/__init__.pxd" */ + +#ifdef _OPENMP +#include +#endif /* _OPENMP */ + +#if defined(PYREX_WITHOUT_ASSERTIONS) && !defined(CYTHON_WITHOUT_ASSERTIONS) +#define CYTHON_WITHOUT_ASSERTIONS +#endif + +typedef struct {PyObject **p; const char *s; const Py_ssize_t n; const char* encoding; + const char is_unicode; const char is_str; const char intern; } __Pyx_StringTabEntry; + +#define __PYX_DEFAULT_STRING_ENCODING_IS_ASCII 0 +#define __PYX_DEFAULT_STRING_ENCODING_IS_UTF8 0 +#define __PYX_DEFAULT_STRING_ENCODING_IS_DEFAULT (PY_MAJOR_VERSION >= 3 && __PYX_DEFAULT_STRING_ENCODING_IS_UTF8) +#define __PYX_DEFAULT_STRING_ENCODING "" +#define __Pyx_PyObject_FromString __Pyx_PyBytes_FromString +#define __Pyx_PyObject_FromStringAndSize __Pyx_PyBytes_FromStringAndSize +#define __Pyx_uchar_cast(c) ((unsigned char)c) +#define __Pyx_long_cast(x) ((long)x) +#define __Pyx_fits_Py_ssize_t(v, type, is_signed) (\ + (sizeof(type) < sizeof(Py_ssize_t)) ||\ + (sizeof(type) > sizeof(Py_ssize_t) &&\ + likely(v < (type)PY_SSIZE_T_MAX ||\ + v == (type)PY_SSIZE_T_MAX) &&\ + (!is_signed || likely(v > (type)PY_SSIZE_T_MIN ||\ + v == (type)PY_SSIZE_T_MIN))) ||\ + (sizeof(type) == sizeof(Py_ssize_t) &&\ + (is_signed || likely(v < (type)PY_SSIZE_T_MAX ||\ + v == (type)PY_SSIZE_T_MAX))) ) +static CYTHON_INLINE int __Pyx_is_valid_index(Py_ssize_t i, Py_ssize_t limit) { + return (size_t) i < (size_t) limit; +} +#if defined (__cplusplus) && __cplusplus >= 201103L + #include + #define __Pyx_sst_abs(value) std::abs(value) +#elif SIZEOF_INT >= SIZEOF_SIZE_T + #define __Pyx_sst_abs(value) abs(value) +#elif SIZEOF_LONG >= SIZEOF_SIZE_T + #define __Pyx_sst_abs(value) labs(value) +#elif defined (_MSC_VER) + #define __Pyx_sst_abs(value) ((Py_ssize_t)_abs64(value)) +#elif defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L + #define __Pyx_sst_abs(value) llabs(value) +#elif defined (__GNUC__) + #define __Pyx_sst_abs(value) __builtin_llabs(value) +#else + #define __Pyx_sst_abs(value) ((value<0) ? -value : value) +#endif +static CYTHON_INLINE const char* __Pyx_PyObject_AsString(PyObject*); +static CYTHON_INLINE const char* __Pyx_PyObject_AsStringAndSize(PyObject*, Py_ssize_t* length); +#define __Pyx_PyByteArray_FromString(s) PyByteArray_FromStringAndSize((const char*)s, strlen((const char*)s)) +#define __Pyx_PyByteArray_FromStringAndSize(s, l) PyByteArray_FromStringAndSize((const char*)s, l) +#define __Pyx_PyBytes_FromString PyBytes_FromString +#define __Pyx_PyBytes_FromStringAndSize PyBytes_FromStringAndSize +static CYTHON_INLINE PyObject* __Pyx_PyUnicode_FromString(const char*); +#if PY_MAJOR_VERSION < 3 + #define __Pyx_PyStr_FromString __Pyx_PyBytes_FromString + #define __Pyx_PyStr_FromStringAndSize __Pyx_PyBytes_FromStringAndSize +#else + #define __Pyx_PyStr_FromString __Pyx_PyUnicode_FromString + #define __Pyx_PyStr_FromStringAndSize __Pyx_PyUnicode_FromStringAndSize +#endif +#define __Pyx_PyBytes_AsWritableString(s) ((char*) PyBytes_AS_STRING(s)) +#define __Pyx_PyBytes_AsWritableSString(s) ((signed char*) PyBytes_AS_STRING(s)) +#define __Pyx_PyBytes_AsWritableUString(s) ((unsigned char*) PyBytes_AS_STRING(s)) +#define __Pyx_PyBytes_AsString(s) ((const char*) PyBytes_AS_STRING(s)) +#define __Pyx_PyBytes_AsSString(s) ((const signed char*) PyBytes_AS_STRING(s)) +#define __Pyx_PyBytes_AsUString(s) ((const unsigned char*) PyBytes_AS_STRING(s)) +#define __Pyx_PyObject_AsWritableString(s) ((char*) __Pyx_PyObject_AsString(s)) +#define __Pyx_PyObject_AsWritableSString(s) ((signed char*) __Pyx_PyObject_AsString(s)) +#define __Pyx_PyObject_AsWritableUString(s) ((unsigned char*) __Pyx_PyObject_AsString(s)) +#define __Pyx_PyObject_AsSString(s) ((const signed char*) __Pyx_PyObject_AsString(s)) +#define __Pyx_PyObject_AsUString(s) ((const unsigned char*) __Pyx_PyObject_AsString(s)) +#define __Pyx_PyObject_FromCString(s) __Pyx_PyObject_FromString((const char*)s) +#define __Pyx_PyBytes_FromCString(s) __Pyx_PyBytes_FromString((const char*)s) +#define __Pyx_PyByteArray_FromCString(s) __Pyx_PyByteArray_FromString((const char*)s) +#define __Pyx_PyStr_FromCString(s) __Pyx_PyStr_FromString((const char*)s) +#define __Pyx_PyUnicode_FromCString(s) __Pyx_PyUnicode_FromString((const char*)s) +static CYTHON_INLINE size_t __Pyx_Py_UNICODE_strlen(const Py_UNICODE *u) { + const Py_UNICODE *u_end = u; + while (*u_end++) ; + return (size_t)(u_end - u - 1); +} +#define __Pyx_PyUnicode_FromUnicode(u) PyUnicode_FromUnicode(u, __Pyx_Py_UNICODE_strlen(u)) +#define __Pyx_PyUnicode_FromUnicodeAndLength PyUnicode_FromUnicode +#define __Pyx_PyUnicode_AsUnicode PyUnicode_AsUnicode +#define __Pyx_NewRef(obj) (Py_INCREF(obj), obj) +#define __Pyx_Owned_Py_None(b) __Pyx_NewRef(Py_None) +static CYTHON_INLINE PyObject * __Pyx_PyBool_FromLong(long b); +static CYTHON_INLINE int __Pyx_PyObject_IsTrue(PyObject*); +static CYTHON_INLINE int __Pyx_PyObject_IsTrueAndDecref(PyObject*); +static CYTHON_INLINE PyObject* __Pyx_PyNumber_IntOrLong(PyObject* x); +#define __Pyx_PySequence_Tuple(obj)\ + (likely(PyTuple_CheckExact(obj)) ? __Pyx_NewRef(obj) : PySequence_Tuple(obj)) +static CYTHON_INLINE Py_ssize_t __Pyx_PyIndex_AsSsize_t(PyObject*); +static CYTHON_INLINE PyObject * __Pyx_PyInt_FromSize_t(size_t); +static CYTHON_INLINE Py_hash_t __Pyx_PyIndex_AsHash_t(PyObject*); +#if CYTHON_ASSUME_SAFE_MACROS +#define __pyx_PyFloat_AsDouble(x) (PyFloat_CheckExact(x) ? PyFloat_AS_DOUBLE(x) : PyFloat_AsDouble(x)) +#else +#define __pyx_PyFloat_AsDouble(x) PyFloat_AsDouble(x) +#endif +#define __pyx_PyFloat_AsFloat(x) ((float) __pyx_PyFloat_AsDouble(x)) +#if PY_MAJOR_VERSION >= 3 +#define __Pyx_PyNumber_Int(x) (PyLong_CheckExact(x) ? __Pyx_NewRef(x) : PyNumber_Long(x)) +#else +#define __Pyx_PyNumber_Int(x) (PyInt_CheckExact(x) ? __Pyx_NewRef(x) : PyNumber_Int(x)) +#endif +#define __Pyx_PyNumber_Float(x) (PyFloat_CheckExact(x) ? __Pyx_NewRef(x) : PyNumber_Float(x)) +#if PY_MAJOR_VERSION < 3 && __PYX_DEFAULT_STRING_ENCODING_IS_ASCII +static int __Pyx_sys_getdefaultencoding_not_ascii; +static int __Pyx_init_sys_getdefaultencoding_params(void) { + PyObject* sys; + PyObject* default_encoding = NULL; + PyObject* ascii_chars_u = NULL; + PyObject* ascii_chars_b = NULL; + const char* default_encoding_c; + sys = PyImport_ImportModule("sys"); + if (!sys) goto bad; + default_encoding = PyObject_CallMethod(sys, (char*) "getdefaultencoding", NULL); + Py_DECREF(sys); + if (!default_encoding) goto bad; + default_encoding_c = PyBytes_AsString(default_encoding); + if (!default_encoding_c) goto bad; + if (strcmp(default_encoding_c, "ascii") == 0) { + __Pyx_sys_getdefaultencoding_not_ascii = 0; + } else { + char ascii_chars[128]; + int c; + for (c = 0; c < 128; c++) { + ascii_chars[c] = c; + } + __Pyx_sys_getdefaultencoding_not_ascii = 1; + ascii_chars_u = PyUnicode_DecodeASCII(ascii_chars, 128, NULL); + if (!ascii_chars_u) goto bad; + ascii_chars_b = PyUnicode_AsEncodedString(ascii_chars_u, default_encoding_c, NULL); + if (!ascii_chars_b || !PyBytes_Check(ascii_chars_b) || memcmp(ascii_chars, PyBytes_AS_STRING(ascii_chars_b), 128) != 0) { + PyErr_Format( + PyExc_ValueError, + "This module compiled with c_string_encoding=ascii, but default encoding '%.200s' is not a superset of ascii.", + default_encoding_c); + goto bad; + } + Py_DECREF(ascii_chars_u); + Py_DECREF(ascii_chars_b); + } + Py_DECREF(default_encoding); + return 0; +bad: + Py_XDECREF(default_encoding); + Py_XDECREF(ascii_chars_u); + Py_XDECREF(ascii_chars_b); + return -1; +} +#endif +#if __PYX_DEFAULT_STRING_ENCODING_IS_DEFAULT && PY_MAJOR_VERSION >= 3 +#define __Pyx_PyUnicode_FromStringAndSize(c_str, size) PyUnicode_DecodeUTF8(c_str, size, NULL) +#else +#define __Pyx_PyUnicode_FromStringAndSize(c_str, size) PyUnicode_Decode(c_str, size, __PYX_DEFAULT_STRING_ENCODING, NULL) +#if __PYX_DEFAULT_STRING_ENCODING_IS_DEFAULT +static char* __PYX_DEFAULT_STRING_ENCODING; +static int __Pyx_init_sys_getdefaultencoding_params(void) { + PyObject* sys; + PyObject* default_encoding = NULL; + char* default_encoding_c; + sys = PyImport_ImportModule("sys"); + if (!sys) goto bad; + default_encoding = PyObject_CallMethod(sys, (char*) (const char*) "getdefaultencoding", NULL); + Py_DECREF(sys); + if (!default_encoding) goto bad; + default_encoding_c = PyBytes_AsString(default_encoding); + if (!default_encoding_c) goto bad; + __PYX_DEFAULT_STRING_ENCODING = (char*) malloc(strlen(default_encoding_c) + 1); + if (!__PYX_DEFAULT_STRING_ENCODING) goto bad; + strcpy(__PYX_DEFAULT_STRING_ENCODING, default_encoding_c); + Py_DECREF(default_encoding); + return 0; +bad: + Py_XDECREF(default_encoding); + return -1; +} +#endif +#endif + + +/* Test for GCC > 2.95 */ +#if defined(__GNUC__) && (__GNUC__ > 2 || (__GNUC__ == 2 && (__GNUC_MINOR__ > 95))) + #define likely(x) __builtin_expect(!!(x), 1) + #define unlikely(x) __builtin_expect(!!(x), 0) +#else /* !__GNUC__ or GCC < 2.95 */ + #define likely(x) (x) + #define unlikely(x) (x) +#endif /* __GNUC__ */ +static CYTHON_INLINE void __Pyx_pretend_to_initialize(void* ptr) { (void)ptr; } + +static PyObject *__pyx_m = NULL; +static PyObject *__pyx_d; +static PyObject *__pyx_b; +static PyObject *__pyx_cython_runtime = NULL; +static PyObject *__pyx_empty_tuple; +static PyObject *__pyx_empty_bytes; +static PyObject *__pyx_empty_unicode; +static int __pyx_lineno; +static int __pyx_clineno = 0; +static const char * __pyx_cfilenm= __FILE__; +static const char *__pyx_filename; + +/* Header.proto */ +#if !defined(CYTHON_CCOMPLEX) + #if defined(__cplusplus) + #define CYTHON_CCOMPLEX 1 + #elif defined(_Complex_I) + #define CYTHON_CCOMPLEX 1 + #else + #define CYTHON_CCOMPLEX 0 + #endif +#endif +#if CYTHON_CCOMPLEX + #ifdef __cplusplus + #include + #else + #include + #endif +#endif +#if CYTHON_CCOMPLEX && !defined(__cplusplus) && defined(__sun__) && defined(__GNUC__) + #undef _Complex_I + #define _Complex_I 1.0fj +#endif + + +static const char *__pyx_f[] = { + "box_overlaps.pyx", + "__init__.pxd", + "type.pxd", +}; +/* BufferFormatStructs.proto */ +#define IS_UNSIGNED(type) (((type) -1) > 0) +struct __Pyx_StructField_; +#define __PYX_BUF_FLAGS_PACKED_STRUCT (1 << 0) +typedef struct { + const char* name; + struct __Pyx_StructField_* fields; + size_t size; + size_t arraysize[8]; + int ndim; + char typegroup; + char is_unsigned; + int flags; +} __Pyx_TypeInfo; +typedef struct __Pyx_StructField_ { + __Pyx_TypeInfo* type; + const char* name; + size_t offset; +} __Pyx_StructField; +typedef struct { + __Pyx_StructField* field; + size_t parent_offset; +} __Pyx_BufFmt_StackElem; +typedef struct { + __Pyx_StructField root; + __Pyx_BufFmt_StackElem* head; + size_t fmt_offset; + size_t new_count, enc_count; + size_t struct_alignment; + int is_complex; + char enc_type; + char new_packmode; + char enc_packmode; + char is_valid_array; +} __Pyx_BufFmt_Context; + + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":689 + * # in Cython to enable them only on the right systems. + * + * ctypedef npy_int8 int8_t # <<<<<<<<<<<<<< + * ctypedef npy_int16 int16_t + * ctypedef npy_int32 int32_t + */ +typedef npy_int8 __pyx_t_5numpy_int8_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":690 + * + * ctypedef npy_int8 int8_t + * ctypedef npy_int16 int16_t # <<<<<<<<<<<<<< + * ctypedef npy_int32 int32_t + * ctypedef npy_int64 int64_t + */ +typedef npy_int16 __pyx_t_5numpy_int16_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":691 + * ctypedef npy_int8 int8_t + * ctypedef npy_int16 int16_t + * ctypedef npy_int32 int32_t # <<<<<<<<<<<<<< + * ctypedef npy_int64 int64_t + * #ctypedef npy_int96 int96_t + */ +typedef npy_int32 __pyx_t_5numpy_int32_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":692 + * ctypedef npy_int16 int16_t + * ctypedef npy_int32 int32_t + * ctypedef npy_int64 int64_t # <<<<<<<<<<<<<< + * #ctypedef npy_int96 int96_t + * #ctypedef npy_int128 int128_t + */ +typedef npy_int64 __pyx_t_5numpy_int64_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":696 + * #ctypedef npy_int128 int128_t + * + * ctypedef npy_uint8 uint8_t # <<<<<<<<<<<<<< + * ctypedef npy_uint16 uint16_t + * ctypedef npy_uint32 uint32_t + */ +typedef npy_uint8 __pyx_t_5numpy_uint8_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":697 + * + * ctypedef npy_uint8 uint8_t + * ctypedef npy_uint16 uint16_t # <<<<<<<<<<<<<< + * ctypedef npy_uint32 uint32_t + * ctypedef npy_uint64 uint64_t + */ +typedef npy_uint16 __pyx_t_5numpy_uint16_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":698 + * ctypedef npy_uint8 uint8_t + * ctypedef npy_uint16 uint16_t + * ctypedef npy_uint32 uint32_t # <<<<<<<<<<<<<< + * ctypedef npy_uint64 uint64_t + * #ctypedef npy_uint96 uint96_t + */ +typedef npy_uint32 __pyx_t_5numpy_uint32_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":699 + * ctypedef npy_uint16 uint16_t + * ctypedef npy_uint32 uint32_t + * ctypedef npy_uint64 uint64_t # <<<<<<<<<<<<<< + * #ctypedef npy_uint96 uint96_t + * #ctypedef npy_uint128 uint128_t + */ +typedef npy_uint64 __pyx_t_5numpy_uint64_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":703 + * #ctypedef npy_uint128 uint128_t + * + * ctypedef npy_float32 float32_t # <<<<<<<<<<<<<< + * ctypedef npy_float64 float64_t + * #ctypedef npy_float80 float80_t + */ +typedef npy_float32 __pyx_t_5numpy_float32_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":704 + * + * ctypedef npy_float32 float32_t + * ctypedef npy_float64 float64_t # <<<<<<<<<<<<<< + * #ctypedef npy_float80 float80_t + * #ctypedef npy_float128 float128_t + */ +typedef npy_float64 __pyx_t_5numpy_float64_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":713 + * # The int types are mapped a bit surprising -- + * # numpy.int corresponds to 'l' and numpy.long to 'q' + * ctypedef npy_long int_t # <<<<<<<<<<<<<< + * ctypedef npy_longlong long_t + * ctypedef npy_longlong longlong_t + */ +typedef npy_long __pyx_t_5numpy_int_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":714 + * # numpy.int corresponds to 'l' and numpy.long to 'q' + * ctypedef npy_long int_t + * ctypedef npy_longlong long_t # <<<<<<<<<<<<<< + * ctypedef npy_longlong longlong_t + * + */ +typedef npy_longlong __pyx_t_5numpy_long_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":715 + * ctypedef npy_long int_t + * ctypedef npy_longlong long_t + * ctypedef npy_longlong longlong_t # <<<<<<<<<<<<<< + * + * ctypedef npy_ulong uint_t + */ +typedef npy_longlong __pyx_t_5numpy_longlong_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":717 + * ctypedef npy_longlong longlong_t + * + * ctypedef npy_ulong uint_t # <<<<<<<<<<<<<< + * ctypedef npy_ulonglong ulong_t + * ctypedef npy_ulonglong ulonglong_t + */ +typedef npy_ulong __pyx_t_5numpy_uint_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":718 + * + * ctypedef npy_ulong uint_t + * ctypedef npy_ulonglong ulong_t # <<<<<<<<<<<<<< + * ctypedef npy_ulonglong ulonglong_t + * + */ +typedef npy_ulonglong __pyx_t_5numpy_ulong_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":719 + * ctypedef npy_ulong uint_t + * ctypedef npy_ulonglong ulong_t + * ctypedef npy_ulonglong ulonglong_t # <<<<<<<<<<<<<< + * + * ctypedef npy_intp intp_t + */ +typedef npy_ulonglong __pyx_t_5numpy_ulonglong_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":721 + * ctypedef npy_ulonglong ulonglong_t + * + * ctypedef npy_intp intp_t # <<<<<<<<<<<<<< + * ctypedef npy_uintp uintp_t + * + */ +typedef npy_intp __pyx_t_5numpy_intp_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":722 + * + * ctypedef npy_intp intp_t + * ctypedef npy_uintp uintp_t # <<<<<<<<<<<<<< + * + * ctypedef npy_double float_t + */ +typedef npy_uintp __pyx_t_5numpy_uintp_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":724 + * ctypedef npy_uintp uintp_t + * + * ctypedef npy_double float_t # <<<<<<<<<<<<<< + * ctypedef npy_double double_t + * ctypedef npy_longdouble longdouble_t + */ +typedef npy_double __pyx_t_5numpy_float_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":725 + * + * ctypedef npy_double float_t + * ctypedef npy_double double_t # <<<<<<<<<<<<<< + * ctypedef npy_longdouble longdouble_t + * + */ +typedef npy_double __pyx_t_5numpy_double_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":726 + * ctypedef npy_double float_t + * ctypedef npy_double double_t + * ctypedef npy_longdouble longdouble_t # <<<<<<<<<<<<<< + * + * ctypedef npy_cfloat cfloat_t + */ +typedef npy_longdouble __pyx_t_5numpy_longdouble_t; + +/* "box_overlaps.pyx":13 + * + * DTYPE = np.float + * ctypedef np.float_t DTYPE_t # <<<<<<<<<<<<<< + * + * def bbox_overlaps( + */ +typedef __pyx_t_5numpy_float_t __pyx_t_4bbox_DTYPE_t; +/* Declarations.proto */ +#if CYTHON_CCOMPLEX + #ifdef __cplusplus + typedef ::std::complex< float > __pyx_t_float_complex; + #else + typedef float _Complex __pyx_t_float_complex; + #endif +#else + typedef struct { float real, imag; } __pyx_t_float_complex; +#endif +static CYTHON_INLINE __pyx_t_float_complex __pyx_t_float_complex_from_parts(float, float); + +/* Declarations.proto */ +#if CYTHON_CCOMPLEX + #ifdef __cplusplus + typedef ::std::complex< double > __pyx_t_double_complex; + #else + typedef double _Complex __pyx_t_double_complex; + #endif +#else + typedef struct { double real, imag; } __pyx_t_double_complex; +#endif +static CYTHON_INLINE __pyx_t_double_complex __pyx_t_double_complex_from_parts(double, double); + + +/*--- Type declarations ---*/ + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":728 + * ctypedef npy_longdouble longdouble_t + * + * ctypedef npy_cfloat cfloat_t # <<<<<<<<<<<<<< + * ctypedef npy_cdouble cdouble_t + * ctypedef npy_clongdouble clongdouble_t + */ +typedef npy_cfloat __pyx_t_5numpy_cfloat_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":729 + * + * ctypedef npy_cfloat cfloat_t + * ctypedef npy_cdouble cdouble_t # <<<<<<<<<<<<<< + * ctypedef npy_clongdouble clongdouble_t + * + */ +typedef npy_cdouble __pyx_t_5numpy_cdouble_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":730 + * ctypedef npy_cfloat cfloat_t + * ctypedef npy_cdouble cdouble_t + * ctypedef npy_clongdouble clongdouble_t # <<<<<<<<<<<<<< + * + * ctypedef npy_cdouble complex_t + */ +typedef npy_clongdouble __pyx_t_5numpy_clongdouble_t; + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":732 + * ctypedef npy_clongdouble clongdouble_t + * + * ctypedef npy_cdouble complex_t # <<<<<<<<<<<<<< + * + * cdef inline object PyArray_MultiIterNew1(a): + */ +typedef npy_cdouble __pyx_t_5numpy_complex_t; + +/* --- Runtime support code (head) --- */ +/* Refnanny.proto */ +#ifndef CYTHON_REFNANNY + #define CYTHON_REFNANNY 0 +#endif +#if CYTHON_REFNANNY + typedef struct { + void (*INCREF)(void*, PyObject*, int); + void (*DECREF)(void*, PyObject*, int); + void (*GOTREF)(void*, PyObject*, int); + void (*GIVEREF)(void*, PyObject*, int); + void* (*SetupContext)(const char*, int, const char*); + void (*FinishContext)(void**); + } __Pyx_RefNannyAPIStruct; + static __Pyx_RefNannyAPIStruct *__Pyx_RefNanny = NULL; + static __Pyx_RefNannyAPIStruct *__Pyx_RefNannyImportAPI(const char *modname); + #define __Pyx_RefNannyDeclarations void *__pyx_refnanny = NULL; +#ifdef WITH_THREAD + #define __Pyx_RefNannySetupContext(name, acquire_gil)\ + if (acquire_gil) {\ + PyGILState_STATE __pyx_gilstate_save = PyGILState_Ensure();\ + __pyx_refnanny = __Pyx_RefNanny->SetupContext((name), __LINE__, __FILE__);\ + PyGILState_Release(__pyx_gilstate_save);\ + } else {\ + __pyx_refnanny = __Pyx_RefNanny->SetupContext((name), __LINE__, __FILE__);\ + } +#else + #define __Pyx_RefNannySetupContext(name, acquire_gil)\ + __pyx_refnanny = __Pyx_RefNanny->SetupContext((name), __LINE__, __FILE__) +#endif + #define __Pyx_RefNannyFinishContext()\ + __Pyx_RefNanny->FinishContext(&__pyx_refnanny) + #define __Pyx_INCREF(r) __Pyx_RefNanny->INCREF(__pyx_refnanny, (PyObject *)(r), __LINE__) + #define __Pyx_DECREF(r) __Pyx_RefNanny->DECREF(__pyx_refnanny, (PyObject *)(r), __LINE__) + #define __Pyx_GOTREF(r) __Pyx_RefNanny->GOTREF(__pyx_refnanny, (PyObject *)(r), __LINE__) + #define __Pyx_GIVEREF(r) __Pyx_RefNanny->GIVEREF(__pyx_refnanny, (PyObject *)(r), __LINE__) + #define __Pyx_XINCREF(r) do { if((r) != NULL) {__Pyx_INCREF(r); }} while(0) + #define __Pyx_XDECREF(r) do { if((r) != NULL) {__Pyx_DECREF(r); }} while(0) + #define __Pyx_XGOTREF(r) do { if((r) != NULL) {__Pyx_GOTREF(r); }} while(0) + #define __Pyx_XGIVEREF(r) do { if((r) != NULL) {__Pyx_GIVEREF(r);}} while(0) +#else + #define __Pyx_RefNannyDeclarations + #define __Pyx_RefNannySetupContext(name, acquire_gil) + #define __Pyx_RefNannyFinishContext() + #define __Pyx_INCREF(r) Py_INCREF(r) + #define __Pyx_DECREF(r) Py_DECREF(r) + #define __Pyx_GOTREF(r) + #define __Pyx_GIVEREF(r) + #define __Pyx_XINCREF(r) Py_XINCREF(r) + #define __Pyx_XDECREF(r) Py_XDECREF(r) + #define __Pyx_XGOTREF(r) + #define __Pyx_XGIVEREF(r) +#endif +#define __Pyx_XDECREF_SET(r, v) do {\ + PyObject *tmp = (PyObject *) r;\ + r = v; __Pyx_XDECREF(tmp);\ + } while (0) +#define __Pyx_DECREF_SET(r, v) do {\ + PyObject *tmp = (PyObject *) r;\ + r = v; __Pyx_DECREF(tmp);\ + } while (0) +#define __Pyx_CLEAR(r) do { PyObject* tmp = ((PyObject*)(r)); r = NULL; __Pyx_DECREF(tmp);} while(0) +#define __Pyx_XCLEAR(r) do { if((r) != NULL) {PyObject* tmp = ((PyObject*)(r)); r = NULL; __Pyx_DECREF(tmp);}} while(0) + +/* PyObjectGetAttrStr.proto */ +#if CYTHON_USE_TYPE_SLOTS +static CYTHON_INLINE PyObject* __Pyx_PyObject_GetAttrStr(PyObject* obj, PyObject* attr_name); +#else +#define __Pyx_PyObject_GetAttrStr(o,n) PyObject_GetAttr(o,n) +#endif + +/* GetBuiltinName.proto */ +static PyObject *__Pyx_GetBuiltinName(PyObject *name); + +/* RaiseArgTupleInvalid.proto */ +static void __Pyx_RaiseArgtupleInvalid(const char* func_name, int exact, + Py_ssize_t num_min, Py_ssize_t num_max, Py_ssize_t num_found); + +/* RaiseDoubleKeywords.proto */ +static void __Pyx_RaiseDoubleKeywordsError(const char* func_name, PyObject* kw_name); + +/* ParseKeywords.proto */ +static int __Pyx_ParseOptionalKeywords(PyObject *kwds, PyObject **argnames[],\ + PyObject *kwds2, PyObject *values[], Py_ssize_t num_pos_args,\ + const char* function_name); + +/* ArgTypeTest.proto */ +#define __Pyx_ArgTypeTest(obj, type, none_allowed, name, exact)\ + ((likely((Py_TYPE(obj) == type) | (none_allowed && (obj == Py_None)))) ? 1 :\ + __Pyx__ArgTypeTest(obj, type, name, exact)) +static int __Pyx__ArgTypeTest(PyObject *obj, PyTypeObject *type, const char *name, int exact); + +/* IsLittleEndian.proto */ +static CYTHON_INLINE int __Pyx_Is_Little_Endian(void); + +/* BufferFormatCheck.proto */ +static const char* __Pyx_BufFmt_CheckString(__Pyx_BufFmt_Context* ctx, const char* ts); +static void __Pyx_BufFmt_Init(__Pyx_BufFmt_Context* ctx, + __Pyx_BufFmt_StackElem* stack, + __Pyx_TypeInfo* type); + +/* BufferGetAndValidate.proto */ +#define __Pyx_GetBufferAndValidate(buf, obj, dtype, flags, nd, cast, stack)\ + ((obj == Py_None || obj == NULL) ?\ + (__Pyx_ZeroBuffer(buf), 0) :\ + __Pyx__GetBufferAndValidate(buf, obj, dtype, flags, nd, cast, stack)) +static int __Pyx__GetBufferAndValidate(Py_buffer* buf, PyObject* obj, + __Pyx_TypeInfo* dtype, int flags, int nd, int cast, __Pyx_BufFmt_StackElem* stack); +static void __Pyx_ZeroBuffer(Py_buffer* buf); +static CYTHON_INLINE void __Pyx_SafeReleaseBuffer(Py_buffer* info); +static Py_ssize_t __Pyx_minusones[] = { -1, -1, -1, -1, -1, -1, -1, -1 }; +static Py_ssize_t __Pyx_zeros[] = { 0, 0, 0, 0, 0, 0, 0, 0 }; + +/* PyDictVersioning.proto */ +#if CYTHON_USE_DICT_VERSIONS && CYTHON_USE_TYPE_SLOTS +#define __PYX_DICT_VERSION_INIT ((PY_UINT64_T) -1) +#define __PYX_GET_DICT_VERSION(dict) (((PyDictObject*)(dict))->ma_version_tag) +#define __PYX_UPDATE_DICT_CACHE(dict, value, cache_var, version_var)\ + (version_var) = __PYX_GET_DICT_VERSION(dict);\ + (cache_var) = (value); +#define __PYX_PY_DICT_LOOKUP_IF_MODIFIED(VAR, DICT, LOOKUP) {\ + static PY_UINT64_T __pyx_dict_version = 0;\ + static PyObject *__pyx_dict_cached_value = NULL;\ + if (likely(__PYX_GET_DICT_VERSION(DICT) == __pyx_dict_version)) {\ + (VAR) = __pyx_dict_cached_value;\ + } else {\ + (VAR) = __pyx_dict_cached_value = (LOOKUP);\ + __pyx_dict_version = __PYX_GET_DICT_VERSION(DICT);\ + }\ +} +static CYTHON_INLINE PY_UINT64_T __Pyx_get_tp_dict_version(PyObject *obj); +static CYTHON_INLINE PY_UINT64_T __Pyx_get_object_dict_version(PyObject *obj); +static CYTHON_INLINE int __Pyx_object_dict_version_matches(PyObject* obj, PY_UINT64_T tp_dict_version, PY_UINT64_T obj_dict_version); +#else +#define __PYX_GET_DICT_VERSION(dict) (0) +#define __PYX_UPDATE_DICT_CACHE(dict, value, cache_var, version_var) +#define __PYX_PY_DICT_LOOKUP_IF_MODIFIED(VAR, DICT, LOOKUP) (VAR) = (LOOKUP); +#endif + +/* GetModuleGlobalName.proto */ +#if CYTHON_USE_DICT_VERSIONS +#define __Pyx_GetModuleGlobalName(var, name) do {\ + static PY_UINT64_T __pyx_dict_version = 0;\ + static PyObject *__pyx_dict_cached_value = NULL;\ + (var) = (likely(__pyx_dict_version == __PYX_GET_DICT_VERSION(__pyx_d))) ?\ + (likely(__pyx_dict_cached_value) ? __Pyx_NewRef(__pyx_dict_cached_value) : __Pyx_GetBuiltinName(name)) :\ + __Pyx__GetModuleGlobalName(name, &__pyx_dict_version, &__pyx_dict_cached_value);\ +} while(0) +#define __Pyx_GetModuleGlobalNameUncached(var, name) do {\ + PY_UINT64_T __pyx_dict_version;\ + PyObject *__pyx_dict_cached_value;\ + (var) = __Pyx__GetModuleGlobalName(name, &__pyx_dict_version, &__pyx_dict_cached_value);\ +} while(0) +static PyObject *__Pyx__GetModuleGlobalName(PyObject *name, PY_UINT64_T *dict_version, PyObject **dict_cached_value); +#else +#define __Pyx_GetModuleGlobalName(var, name) (var) = __Pyx__GetModuleGlobalName(name) +#define __Pyx_GetModuleGlobalNameUncached(var, name) (var) = __Pyx__GetModuleGlobalName(name) +static CYTHON_INLINE PyObject *__Pyx__GetModuleGlobalName(PyObject *name); +#endif + +/* PyObjectCall.proto */ +#if CYTHON_COMPILING_IN_CPYTHON +static CYTHON_INLINE PyObject* __Pyx_PyObject_Call(PyObject *func, PyObject *arg, PyObject *kw); +#else +#define __Pyx_PyObject_Call(func, arg, kw) PyObject_Call(func, arg, kw) +#endif + +/* ExtTypeTest.proto */ +static CYTHON_INLINE int __Pyx_TypeTest(PyObject *obj, PyTypeObject *type); + +/* BufferIndexError.proto */ +static void __Pyx_RaiseBufferIndexError(int axis); + +#define __Pyx_BufPtrStrided2d(type, buf, i0, s0, i1, s1) (type)((char*)buf + i0 * s0 + i1 * s1) +/* PyThreadStateGet.proto */ +#if CYTHON_FAST_THREAD_STATE +#define __Pyx_PyThreadState_declare PyThreadState *__pyx_tstate; +#define __Pyx_PyThreadState_assign __pyx_tstate = __Pyx_PyThreadState_Current; +#define __Pyx_PyErr_Occurred() __pyx_tstate->curexc_type +#else +#define __Pyx_PyThreadState_declare +#define __Pyx_PyThreadState_assign +#define __Pyx_PyErr_Occurred() PyErr_Occurred() +#endif + +/* PyErrFetchRestore.proto */ +#if CYTHON_FAST_THREAD_STATE +#define __Pyx_PyErr_Clear() __Pyx_ErrRestore(NULL, NULL, NULL) +#define __Pyx_ErrRestoreWithState(type, value, tb) __Pyx_ErrRestoreInState(PyThreadState_GET(), type, value, tb) +#define __Pyx_ErrFetchWithState(type, value, tb) __Pyx_ErrFetchInState(PyThreadState_GET(), type, value, tb) +#define __Pyx_ErrRestore(type, value, tb) __Pyx_ErrRestoreInState(__pyx_tstate, type, value, tb) +#define __Pyx_ErrFetch(type, value, tb) __Pyx_ErrFetchInState(__pyx_tstate, type, value, tb) +static CYTHON_INLINE void __Pyx_ErrRestoreInState(PyThreadState *tstate, PyObject *type, PyObject *value, PyObject *tb); +static CYTHON_INLINE void __Pyx_ErrFetchInState(PyThreadState *tstate, PyObject **type, PyObject **value, PyObject **tb); +#if CYTHON_COMPILING_IN_CPYTHON +#define __Pyx_PyErr_SetNone(exc) (Py_INCREF(exc), __Pyx_ErrRestore((exc), NULL, NULL)) +#else +#define __Pyx_PyErr_SetNone(exc) PyErr_SetNone(exc) +#endif +#else +#define __Pyx_PyErr_Clear() PyErr_Clear() +#define __Pyx_PyErr_SetNone(exc) PyErr_SetNone(exc) +#define __Pyx_ErrRestoreWithState(type, value, tb) PyErr_Restore(type, value, tb) +#define __Pyx_ErrFetchWithState(type, value, tb) PyErr_Fetch(type, value, tb) +#define __Pyx_ErrRestoreInState(tstate, type, value, tb) PyErr_Restore(type, value, tb) +#define __Pyx_ErrFetchInState(tstate, type, value, tb) PyErr_Fetch(type, value, tb) +#define __Pyx_ErrRestore(type, value, tb) PyErr_Restore(type, value, tb) +#define __Pyx_ErrFetch(type, value, tb) PyErr_Fetch(type, value, tb) +#endif + +/* GetTopmostException.proto */ +#if CYTHON_USE_EXC_INFO_STACK +static _PyErr_StackItem * __Pyx_PyErr_GetTopmostException(PyThreadState *tstate); +#endif + +/* SaveResetException.proto */ +#if CYTHON_FAST_THREAD_STATE +#define __Pyx_ExceptionSave(type, value, tb) __Pyx__ExceptionSave(__pyx_tstate, type, value, tb) +static CYTHON_INLINE void __Pyx__ExceptionSave(PyThreadState *tstate, PyObject **type, PyObject **value, PyObject **tb); +#define __Pyx_ExceptionReset(type, value, tb) __Pyx__ExceptionReset(__pyx_tstate, type, value, tb) +static CYTHON_INLINE void __Pyx__ExceptionReset(PyThreadState *tstate, PyObject *type, PyObject *value, PyObject *tb); +#else +#define __Pyx_ExceptionSave(type, value, tb) PyErr_GetExcInfo(type, value, tb) +#define __Pyx_ExceptionReset(type, value, tb) PyErr_SetExcInfo(type, value, tb) +#endif + +/* PyErrExceptionMatches.proto */ +#if CYTHON_FAST_THREAD_STATE +#define __Pyx_PyErr_ExceptionMatches(err) __Pyx_PyErr_ExceptionMatchesInState(__pyx_tstate, err) +static CYTHON_INLINE int __Pyx_PyErr_ExceptionMatchesInState(PyThreadState* tstate, PyObject* err); +#else +#define __Pyx_PyErr_ExceptionMatches(err) PyErr_ExceptionMatches(err) +#endif + +/* GetException.proto */ +#if CYTHON_FAST_THREAD_STATE +#define __Pyx_GetException(type, value, tb) __Pyx__GetException(__pyx_tstate, type, value, tb) +static int __Pyx__GetException(PyThreadState *tstate, PyObject **type, PyObject **value, PyObject **tb); +#else +static int __Pyx_GetException(PyObject **type, PyObject **value, PyObject **tb); +#endif + +/* RaiseException.proto */ +static void __Pyx_Raise(PyObject *type, PyObject *value, PyObject *tb, PyObject *cause); + +/* TypeImport.proto */ +#ifndef __PYX_HAVE_RT_ImportType_proto_0_29_36 +#define __PYX_HAVE_RT_ImportType_proto_0_29_36 +#if __STDC_VERSION__ >= 201112L +#include +#endif +#if __STDC_VERSION__ >= 201112L || __cplusplus >= 201103L +#define __PYX_GET_STRUCT_ALIGNMENT_0_29_36(s) alignof(s) +#else +#define __PYX_GET_STRUCT_ALIGNMENT_0_29_36(s) sizeof(void*) +#endif +enum __Pyx_ImportType_CheckSize_0_29_36 { + __Pyx_ImportType_CheckSize_Error_0_29_36 = 0, + __Pyx_ImportType_CheckSize_Warn_0_29_36 = 1, + __Pyx_ImportType_CheckSize_Ignore_0_29_36 = 2 +}; +static PyTypeObject *__Pyx_ImportType_0_29_36(PyObject* module, const char *module_name, const char *class_name, size_t size, size_t alignment, enum __Pyx_ImportType_CheckSize_0_29_36 check_size); +#endif + +/* Import.proto */ +static PyObject *__Pyx_Import(PyObject *name, PyObject *from_list, int level); + +/* CLineInTraceback.proto */ +#ifdef CYTHON_CLINE_IN_TRACEBACK +#define __Pyx_CLineForTraceback(tstate, c_line) (((CYTHON_CLINE_IN_TRACEBACK)) ? c_line : 0) +#else +static int __Pyx_CLineForTraceback(PyThreadState *tstate, int c_line); +#endif + +/* CodeObjectCache.proto */ +typedef struct { + PyCodeObject* code_object; + int code_line; +} __Pyx_CodeObjectCacheEntry; +struct __Pyx_CodeObjectCache { + int count; + int max_count; + __Pyx_CodeObjectCacheEntry* entries; +}; +static struct __Pyx_CodeObjectCache __pyx_code_cache = {0,0,NULL}; +static int __pyx_bisect_code_objects(__Pyx_CodeObjectCacheEntry* entries, int count, int code_line); +static PyCodeObject *__pyx_find_code_object(int code_line); +static void __pyx_insert_code_object(int code_line, PyCodeObject* code_object); + +/* AddTraceback.proto */ +static void __Pyx_AddTraceback(const char *funcname, int c_line, + int py_line, const char *filename); + +/* BufferStructDeclare.proto */ +typedef struct { + Py_ssize_t shape, strides, suboffsets; +} __Pyx_Buf_DimInfo; +typedef struct { + size_t refcount; + Py_buffer pybuffer; +} __Pyx_Buffer; +typedef struct { + __Pyx_Buffer *rcbuffer; + char *data; + __Pyx_Buf_DimInfo diminfo[8]; +} __Pyx_LocalBuf_ND; + +#if PY_MAJOR_VERSION < 3 + static int __Pyx_GetBuffer(PyObject *obj, Py_buffer *view, int flags); + static void __Pyx_ReleaseBuffer(Py_buffer *view); +#else + #define __Pyx_GetBuffer PyObject_GetBuffer + #define __Pyx_ReleaseBuffer PyBuffer_Release +#endif + + +/* GCCDiagnostics.proto */ +#if defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)) +#define __Pyx_HAS_GCC_DIAGNOSTIC +#endif + +/* RealImag.proto */ +#if CYTHON_CCOMPLEX + #ifdef __cplusplus + #define __Pyx_CREAL(z) ((z).real()) + #define __Pyx_CIMAG(z) ((z).imag()) + #else + #define __Pyx_CREAL(z) (__real__(z)) + #define __Pyx_CIMAG(z) (__imag__(z)) + #endif +#else + #define __Pyx_CREAL(z) ((z).real) + #define __Pyx_CIMAG(z) ((z).imag) +#endif +#if defined(__cplusplus) && CYTHON_CCOMPLEX\ + && (defined(_WIN32) || defined(__clang__) || (defined(__GNUC__) && (__GNUC__ >= 5 || __GNUC__ == 4 && __GNUC_MINOR__ >= 4 )) || __cplusplus >= 201103) + #define __Pyx_SET_CREAL(z,x) ((z).real(x)) + #define __Pyx_SET_CIMAG(z,y) ((z).imag(y)) +#else + #define __Pyx_SET_CREAL(z,x) __Pyx_CREAL(z) = (x) + #define __Pyx_SET_CIMAG(z,y) __Pyx_CIMAG(z) = (y) +#endif + +/* Arithmetic.proto */ +#if CYTHON_CCOMPLEX + #define __Pyx_c_eq_float(a, b) ((a)==(b)) + #define __Pyx_c_sum_float(a, b) ((a)+(b)) + #define __Pyx_c_diff_float(a, b) ((a)-(b)) + #define __Pyx_c_prod_float(a, b) ((a)*(b)) + #define __Pyx_c_quot_float(a, b) ((a)/(b)) + #define __Pyx_c_neg_float(a) (-(a)) + #ifdef __cplusplus + #define __Pyx_c_is_zero_float(z) ((z)==(float)0) + #define __Pyx_c_conj_float(z) (::std::conj(z)) + #if 1 + #define __Pyx_c_abs_float(z) (::std::abs(z)) + #define __Pyx_c_pow_float(a, b) (::std::pow(a, b)) + #endif + #else + #define __Pyx_c_is_zero_float(z) ((z)==0) + #define __Pyx_c_conj_float(z) (conjf(z)) + #if 1 + #define __Pyx_c_abs_float(z) (cabsf(z)) + #define __Pyx_c_pow_float(a, b) (cpowf(a, b)) + #endif + #endif +#else + static CYTHON_INLINE int __Pyx_c_eq_float(__pyx_t_float_complex, __pyx_t_float_complex); + static CYTHON_INLINE __pyx_t_float_complex __Pyx_c_sum_float(__pyx_t_float_complex, __pyx_t_float_complex); + static CYTHON_INLINE __pyx_t_float_complex __Pyx_c_diff_float(__pyx_t_float_complex, __pyx_t_float_complex); + static CYTHON_INLINE __pyx_t_float_complex __Pyx_c_prod_float(__pyx_t_float_complex, __pyx_t_float_complex); + static CYTHON_INLINE __pyx_t_float_complex __Pyx_c_quot_float(__pyx_t_float_complex, __pyx_t_float_complex); + static CYTHON_INLINE __pyx_t_float_complex __Pyx_c_neg_float(__pyx_t_float_complex); + static CYTHON_INLINE int __Pyx_c_is_zero_float(__pyx_t_float_complex); + static CYTHON_INLINE __pyx_t_float_complex __Pyx_c_conj_float(__pyx_t_float_complex); + #if 1 + static CYTHON_INLINE float __Pyx_c_abs_float(__pyx_t_float_complex); + static CYTHON_INLINE __pyx_t_float_complex __Pyx_c_pow_float(__pyx_t_float_complex, __pyx_t_float_complex); + #endif +#endif + +/* Arithmetic.proto */ +#if CYTHON_CCOMPLEX + #define __Pyx_c_eq_double(a, b) ((a)==(b)) + #define __Pyx_c_sum_double(a, b) ((a)+(b)) + #define __Pyx_c_diff_double(a, b) ((a)-(b)) + #define __Pyx_c_prod_double(a, b) ((a)*(b)) + #define __Pyx_c_quot_double(a, b) ((a)/(b)) + #define __Pyx_c_neg_double(a) (-(a)) + #ifdef __cplusplus + #define __Pyx_c_is_zero_double(z) ((z)==(double)0) + #define __Pyx_c_conj_double(z) (::std::conj(z)) + #if 1 + #define __Pyx_c_abs_double(z) (::std::abs(z)) + #define __Pyx_c_pow_double(a, b) (::std::pow(a, b)) + #endif + #else + #define __Pyx_c_is_zero_double(z) ((z)==0) + #define __Pyx_c_conj_double(z) (conj(z)) + #if 1 + #define __Pyx_c_abs_double(z) (cabs(z)) + #define __Pyx_c_pow_double(a, b) (cpow(a, b)) + #endif + #endif +#else + static CYTHON_INLINE int __Pyx_c_eq_double(__pyx_t_double_complex, __pyx_t_double_complex); + static CYTHON_INLINE __pyx_t_double_complex __Pyx_c_sum_double(__pyx_t_double_complex, __pyx_t_double_complex); + static CYTHON_INLINE __pyx_t_double_complex __Pyx_c_diff_double(__pyx_t_double_complex, __pyx_t_double_complex); + static CYTHON_INLINE __pyx_t_double_complex __Pyx_c_prod_double(__pyx_t_double_complex, __pyx_t_double_complex); + static CYTHON_INLINE __pyx_t_double_complex __Pyx_c_quot_double(__pyx_t_double_complex, __pyx_t_double_complex); + static CYTHON_INLINE __pyx_t_double_complex __Pyx_c_neg_double(__pyx_t_double_complex); + static CYTHON_INLINE int __Pyx_c_is_zero_double(__pyx_t_double_complex); + static CYTHON_INLINE __pyx_t_double_complex __Pyx_c_conj_double(__pyx_t_double_complex); + #if 1 + static CYTHON_INLINE double __Pyx_c_abs_double(__pyx_t_double_complex); + static CYTHON_INLINE __pyx_t_double_complex __Pyx_c_pow_double(__pyx_t_double_complex, __pyx_t_double_complex); + #endif +#endif + +/* CIntToPy.proto */ +static CYTHON_INLINE PyObject* __Pyx_PyInt_From_unsigned_int(unsigned int value); + +/* CIntFromPy.proto */ +static CYTHON_INLINE unsigned int __Pyx_PyInt_As_unsigned_int(PyObject *); + +/* CIntToPy.proto */ +static CYTHON_INLINE PyObject* __Pyx_PyInt_From_long(long value); + +/* CIntFromPy.proto */ +static CYTHON_INLINE long __Pyx_PyInt_As_long(PyObject *); + +/* CIntFromPy.proto */ +static CYTHON_INLINE int __Pyx_PyInt_As_int(PyObject *); + +/* FastTypeChecks.proto */ +#if CYTHON_COMPILING_IN_CPYTHON +#define __Pyx_TypeCheck(obj, type) __Pyx_IsSubtype(Py_TYPE(obj), (PyTypeObject *)type) +static CYTHON_INLINE int __Pyx_IsSubtype(PyTypeObject *a, PyTypeObject *b); +static CYTHON_INLINE int __Pyx_PyErr_GivenExceptionMatches(PyObject *err, PyObject *type); +static CYTHON_INLINE int __Pyx_PyErr_GivenExceptionMatches2(PyObject *err, PyObject *type1, PyObject *type2); +#else +#define __Pyx_TypeCheck(obj, type) PyObject_TypeCheck(obj, (PyTypeObject *)type) +#define __Pyx_PyErr_GivenExceptionMatches(err, type) PyErr_GivenExceptionMatches(err, type) +#define __Pyx_PyErr_GivenExceptionMatches2(err, type1, type2) (PyErr_GivenExceptionMatches(err, type1) || PyErr_GivenExceptionMatches(err, type2)) +#endif +#define __Pyx_PyException_Check(obj) __Pyx_TypeCheck(obj, PyExc_Exception) + +/* CheckBinaryVersion.proto */ +static int __Pyx_check_binary_version(void); + +/* InitStrings.proto */ +static int __Pyx_InitStrings(__Pyx_StringTabEntry *t); + + +/* Module declarations from 'cython' */ + +/* Module declarations from 'cpython.buffer' */ + +/* Module declarations from 'libc.string' */ + +/* Module declarations from 'libc.stdio' */ + +/* Module declarations from '__builtin__' */ + +/* Module declarations from 'cpython.type' */ +static PyTypeObject *__pyx_ptype_7cpython_4type_type = 0; + +/* Module declarations from 'cpython' */ + +/* Module declarations from 'cpython.object' */ + +/* Module declarations from 'cpython.ref' */ + +/* Module declarations from 'cpython.mem' */ + +/* Module declarations from 'numpy' */ + +/* Module declarations from 'numpy' */ +static PyTypeObject *__pyx_ptype_5numpy_dtype = 0; +static PyTypeObject *__pyx_ptype_5numpy_flatiter = 0; +static PyTypeObject *__pyx_ptype_5numpy_broadcast = 0; +static PyTypeObject *__pyx_ptype_5numpy_ndarray = 0; +static PyTypeObject *__pyx_ptype_5numpy_generic = 0; +static PyTypeObject *__pyx_ptype_5numpy_number = 0; +static PyTypeObject *__pyx_ptype_5numpy_integer = 0; +static PyTypeObject *__pyx_ptype_5numpy_signedinteger = 0; +static PyTypeObject *__pyx_ptype_5numpy_unsignedinteger = 0; +static PyTypeObject *__pyx_ptype_5numpy_inexact = 0; +static PyTypeObject *__pyx_ptype_5numpy_floating = 0; +static PyTypeObject *__pyx_ptype_5numpy_complexfloating = 0; +static PyTypeObject *__pyx_ptype_5numpy_flexible = 0; +static PyTypeObject *__pyx_ptype_5numpy_character = 0; +static PyTypeObject *__pyx_ptype_5numpy_ufunc = 0; + +/* Module declarations from 'bbox' */ +static __Pyx_TypeInfo __Pyx_TypeInfo_nn___pyx_t_4bbox_DTYPE_t = { "DTYPE_t", NULL, sizeof(__pyx_t_4bbox_DTYPE_t), { 0 }, 0, 'R', 0, 0 }; +#define __Pyx_MODULE_NAME "bbox" +extern int __pyx_module_is_main_bbox; +int __pyx_module_is_main_bbox = 0; + +/* Implementation of 'bbox' */ +static PyObject *__pyx_builtin_range; +static PyObject *__pyx_builtin_ImportError; +static const char __pyx_k_K[] = "K"; +static const char __pyx_k_N[] = "N"; +static const char __pyx_k_k[] = "k"; +static const char __pyx_k_n[] = "n"; +static const char __pyx_k_ih[] = "ih"; +static const char __pyx_k_iw[] = "iw"; +static const char __pyx_k_np[] = "np"; +static const char __pyx_k_ua[] = "ua"; +static const char __pyx_k_bbox[] = "bbox"; +static const char __pyx_k_main[] = "__main__"; +static const char __pyx_k_name[] = "__name__"; +static const char __pyx_k_test[] = "__test__"; +static const char __pyx_k_DTYPE[] = "DTYPE"; +static const char __pyx_k_boxes[] = "boxes"; +static const char __pyx_k_dtype[] = "dtype"; +static const char __pyx_k_float[] = "float"; +static const char __pyx_k_numpy[] = "numpy"; +static const char __pyx_k_range[] = "range"; +static const char __pyx_k_zeros[] = "zeros"; +static const char __pyx_k_import[] = "__import__"; +static const char __pyx_k_box_area[] = "box_area"; +static const char __pyx_k_overlaps[] = "overlaps"; +static const char __pyx_k_ImportError[] = "ImportError"; +static const char __pyx_k_query_boxes[] = "query_boxes"; +static const char __pyx_k_bbox_overlaps[] = "bbox_overlaps"; +static const char __pyx_k_box_overlaps_pyx[] = "box_overlaps.pyx"; +static const char __pyx_k_cline_in_traceback[] = "cline_in_traceback"; +static const char __pyx_k_numpy_core_multiarray_failed_to[] = "numpy.core.multiarray failed to import"; +static const char __pyx_k_numpy_core_umath_failed_to_impor[] = "numpy.core.umath failed to import"; +static PyObject *__pyx_n_s_DTYPE; +static PyObject *__pyx_n_s_ImportError; +static PyObject *__pyx_n_s_K; +static PyObject *__pyx_n_s_N; +static PyObject *__pyx_n_s_bbox; +static PyObject *__pyx_n_s_bbox_overlaps; +static PyObject *__pyx_n_s_box_area; +static PyObject *__pyx_kp_s_box_overlaps_pyx; +static PyObject *__pyx_n_s_boxes; +static PyObject *__pyx_n_s_cline_in_traceback; +static PyObject *__pyx_n_s_dtype; +static PyObject *__pyx_n_s_float; +static PyObject *__pyx_n_s_ih; +static PyObject *__pyx_n_s_import; +static PyObject *__pyx_n_s_iw; +static PyObject *__pyx_n_s_k; +static PyObject *__pyx_n_s_main; +static PyObject *__pyx_n_s_n; +static PyObject *__pyx_n_s_name; +static PyObject *__pyx_n_s_np; +static PyObject *__pyx_n_s_numpy; +static PyObject *__pyx_kp_s_numpy_core_multiarray_failed_to; +static PyObject *__pyx_kp_s_numpy_core_umath_failed_to_impor; +static PyObject *__pyx_n_s_overlaps; +static PyObject *__pyx_n_s_query_boxes; +static PyObject *__pyx_n_s_range; +static PyObject *__pyx_n_s_test; +static PyObject *__pyx_n_s_ua; +static PyObject *__pyx_n_s_zeros; +static PyObject *__pyx_pf_4bbox_bbox_overlaps(CYTHON_UNUSED PyObject *__pyx_self, PyArrayObject *__pyx_v_boxes, PyArrayObject *__pyx_v_query_boxes); /* proto */ +static PyObject *__pyx_tuple_; +static PyObject *__pyx_tuple__2; +static PyObject *__pyx_tuple__3; +static PyObject *__pyx_codeobj__4; +/* Late includes */ + +/* "box_overlaps.pyx":15 + * ctypedef np.float_t DTYPE_t + * + * def bbox_overlaps( # <<<<<<<<<<<<<< + * np.ndarray[DTYPE_t, ndim=2] boxes, + * np.ndarray[DTYPE_t, ndim=2] query_boxes): + */ + +/* Python wrapper */ +static PyObject *__pyx_pw_4bbox_1bbox_overlaps(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/ +static char __pyx_doc_4bbox_bbox_overlaps[] = "\n Parameters\n ----------\n boxes: (N, 4) ndarray of float\n query_boxes: (K, 4) ndarray of float\n Returns\n -------\n overlaps: (N, K) ndarray of overlap between boxes and query_boxes\n "; +static PyMethodDef __pyx_mdef_4bbox_1bbox_overlaps = {"bbox_overlaps", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_4bbox_1bbox_overlaps, METH_VARARGS|METH_KEYWORDS, __pyx_doc_4bbox_bbox_overlaps}; +static PyObject *__pyx_pw_4bbox_1bbox_overlaps(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) { + PyArrayObject *__pyx_v_boxes = 0; + PyArrayObject *__pyx_v_query_boxes = 0; + int __pyx_lineno = 0; + const char *__pyx_filename = NULL; + int __pyx_clineno = 0; + PyObject *__pyx_r = 0; + __Pyx_RefNannyDeclarations + __Pyx_RefNannySetupContext("bbox_overlaps (wrapper)", 0); + { + static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_boxes,&__pyx_n_s_query_boxes,0}; + PyObject* values[2] = {0,0}; + if (unlikely(__pyx_kwds)) { + Py_ssize_t kw_args; + const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args); + switch (pos_args) { + case 2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1); + CYTHON_FALLTHROUGH; + case 1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0); + CYTHON_FALLTHROUGH; + case 0: break; + default: goto __pyx_L5_argtuple_error; + } + kw_args = PyDict_Size(__pyx_kwds); + switch (pos_args) { + case 0: + if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_boxes)) != 0)) kw_args--; + else goto __pyx_L5_argtuple_error; + CYTHON_FALLTHROUGH; + case 1: + if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_query_boxes)) != 0)) kw_args--; + else { + __Pyx_RaiseArgtupleInvalid("bbox_overlaps", 1, 2, 2, 1); __PYX_ERR(0, 15, __pyx_L3_error) + } + } + if (unlikely(kw_args > 0)) { + if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "bbox_overlaps") < 0)) __PYX_ERR(0, 15, __pyx_L3_error) + } + } else if (PyTuple_GET_SIZE(__pyx_args) != 2) { + goto __pyx_L5_argtuple_error; + } else { + values[0] = PyTuple_GET_ITEM(__pyx_args, 0); + values[1] = PyTuple_GET_ITEM(__pyx_args, 1); + } + __pyx_v_boxes = ((PyArrayObject *)values[0]); + __pyx_v_query_boxes = ((PyArrayObject *)values[1]); + } + goto __pyx_L4_argument_unpacking_done; + __pyx_L5_argtuple_error:; + __Pyx_RaiseArgtupleInvalid("bbox_overlaps", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 15, __pyx_L3_error) + __pyx_L3_error:; + __Pyx_AddTraceback("bbox.bbox_overlaps", __pyx_clineno, __pyx_lineno, __pyx_filename); + __Pyx_RefNannyFinishContext(); + return NULL; + __pyx_L4_argument_unpacking_done:; + if (unlikely(!__Pyx_ArgTypeTest(((PyObject *)__pyx_v_boxes), __pyx_ptype_5numpy_ndarray, 1, "boxes", 0))) __PYX_ERR(0, 16, __pyx_L1_error) + if (unlikely(!__Pyx_ArgTypeTest(((PyObject *)__pyx_v_query_boxes), __pyx_ptype_5numpy_ndarray, 1, "query_boxes", 0))) __PYX_ERR(0, 17, __pyx_L1_error) + __pyx_r = __pyx_pf_4bbox_bbox_overlaps(__pyx_self, __pyx_v_boxes, __pyx_v_query_boxes); + + /* function exit code */ + goto __pyx_L0; + __pyx_L1_error:; + __pyx_r = NULL; + __pyx_L0:; + __Pyx_RefNannyFinishContext(); + return __pyx_r; +} + +static PyObject *__pyx_pf_4bbox_bbox_overlaps(CYTHON_UNUSED PyObject *__pyx_self, PyArrayObject *__pyx_v_boxes, PyArrayObject *__pyx_v_query_boxes) { + unsigned int __pyx_v_N; + unsigned int __pyx_v_K; + PyArrayObject *__pyx_v_overlaps = 0; + __pyx_t_4bbox_DTYPE_t __pyx_v_iw; + __pyx_t_4bbox_DTYPE_t __pyx_v_ih; + __pyx_t_4bbox_DTYPE_t __pyx_v_box_area; + __pyx_t_4bbox_DTYPE_t __pyx_v_ua; + unsigned int __pyx_v_k; + unsigned int __pyx_v_n; + __Pyx_LocalBuf_ND __pyx_pybuffernd_boxes; + __Pyx_Buffer __pyx_pybuffer_boxes; + __Pyx_LocalBuf_ND __pyx_pybuffernd_overlaps; + __Pyx_Buffer __pyx_pybuffer_overlaps; + __Pyx_LocalBuf_ND __pyx_pybuffernd_query_boxes; + __Pyx_Buffer __pyx_pybuffer_query_boxes; + PyObject *__pyx_r = NULL; + __Pyx_RefNannyDeclarations + PyObject *__pyx_t_1 = NULL; + PyObject *__pyx_t_2 = NULL; + PyObject *__pyx_t_3 = NULL; + PyObject *__pyx_t_4 = NULL; + PyArrayObject *__pyx_t_5 = NULL; + unsigned int __pyx_t_6; + unsigned int __pyx_t_7; + unsigned int __pyx_t_8; + size_t __pyx_t_9; + Py_ssize_t __pyx_t_10; + int __pyx_t_11; + size_t __pyx_t_12; + Py_ssize_t __pyx_t_13; + size_t __pyx_t_14; + Py_ssize_t __pyx_t_15; + size_t __pyx_t_16; + Py_ssize_t __pyx_t_17; + unsigned int __pyx_t_18; + unsigned int __pyx_t_19; + unsigned int __pyx_t_20; + __pyx_t_4bbox_DTYPE_t __pyx_t_21; + __pyx_t_4bbox_DTYPE_t __pyx_t_22; + __pyx_t_4bbox_DTYPE_t __pyx_t_23; + __pyx_t_4bbox_DTYPE_t __pyx_t_24; + int __pyx_t_25; + int __pyx_lineno = 0; + const char *__pyx_filename = NULL; + int __pyx_clineno = 0; + __Pyx_RefNannySetupContext("bbox_overlaps", 0); + __pyx_pybuffer_overlaps.pybuffer.buf = NULL; + __pyx_pybuffer_overlaps.refcount = 0; + __pyx_pybuffernd_overlaps.data = NULL; + __pyx_pybuffernd_overlaps.rcbuffer = &__pyx_pybuffer_overlaps; + __pyx_pybuffer_boxes.pybuffer.buf = NULL; + __pyx_pybuffer_boxes.refcount = 0; + __pyx_pybuffernd_boxes.data = NULL; + __pyx_pybuffernd_boxes.rcbuffer = &__pyx_pybuffer_boxes; + __pyx_pybuffer_query_boxes.pybuffer.buf = NULL; + __pyx_pybuffer_query_boxes.refcount = 0; + __pyx_pybuffernd_query_boxes.data = NULL; + __pyx_pybuffernd_query_boxes.rcbuffer = &__pyx_pybuffer_query_boxes; + { + __Pyx_BufFmt_StackElem __pyx_stack[1]; + if (unlikely(__Pyx_GetBufferAndValidate(&__pyx_pybuffernd_boxes.rcbuffer->pybuffer, (PyObject*)__pyx_v_boxes, &__Pyx_TypeInfo_nn___pyx_t_4bbox_DTYPE_t, PyBUF_FORMAT| PyBUF_STRIDES, 2, 0, __pyx_stack) == -1)) __PYX_ERR(0, 15, __pyx_L1_error) + } + __pyx_pybuffernd_boxes.diminfo[0].strides = __pyx_pybuffernd_boxes.rcbuffer->pybuffer.strides[0]; __pyx_pybuffernd_boxes.diminfo[0].shape = __pyx_pybuffernd_boxes.rcbuffer->pybuffer.shape[0]; __pyx_pybuffernd_boxes.diminfo[1].strides = __pyx_pybuffernd_boxes.rcbuffer->pybuffer.strides[1]; __pyx_pybuffernd_boxes.diminfo[1].shape = __pyx_pybuffernd_boxes.rcbuffer->pybuffer.shape[1]; + { + __Pyx_BufFmt_StackElem __pyx_stack[1]; + if (unlikely(__Pyx_GetBufferAndValidate(&__pyx_pybuffernd_query_boxes.rcbuffer->pybuffer, (PyObject*)__pyx_v_query_boxes, &__Pyx_TypeInfo_nn___pyx_t_4bbox_DTYPE_t, PyBUF_FORMAT| PyBUF_STRIDES, 2, 0, __pyx_stack) == -1)) __PYX_ERR(0, 15, __pyx_L1_error) + } + __pyx_pybuffernd_query_boxes.diminfo[0].strides = __pyx_pybuffernd_query_boxes.rcbuffer->pybuffer.strides[0]; __pyx_pybuffernd_query_boxes.diminfo[0].shape = __pyx_pybuffernd_query_boxes.rcbuffer->pybuffer.shape[0]; __pyx_pybuffernd_query_boxes.diminfo[1].strides = __pyx_pybuffernd_query_boxes.rcbuffer->pybuffer.strides[1]; __pyx_pybuffernd_query_boxes.diminfo[1].shape = __pyx_pybuffernd_query_boxes.rcbuffer->pybuffer.shape[1]; + + /* "box_overlaps.pyx":27 + * overlaps: (N, K) ndarray of overlap between boxes and query_boxes + * """ + * cdef unsigned int N = boxes.shape[0] # <<<<<<<<<<<<<< + * cdef unsigned int K = query_boxes.shape[0] + * cdef np.ndarray[DTYPE_t, ndim=2] overlaps = np.zeros((N, K), dtype=DTYPE) + */ + __pyx_v_N = (__pyx_v_boxes->dimensions[0]); + + /* "box_overlaps.pyx":28 + * """ + * cdef unsigned int N = boxes.shape[0] + * cdef unsigned int K = query_boxes.shape[0] # <<<<<<<<<<<<<< + * cdef np.ndarray[DTYPE_t, ndim=2] overlaps = np.zeros((N, K), dtype=DTYPE) + * cdef DTYPE_t iw, ih, box_area + */ + __pyx_v_K = (__pyx_v_query_boxes->dimensions[0]); + + /* "box_overlaps.pyx":29 + * cdef unsigned int N = boxes.shape[0] + * cdef unsigned int K = query_boxes.shape[0] + * cdef np.ndarray[DTYPE_t, ndim=2] overlaps = np.zeros((N, K), dtype=DTYPE) # <<<<<<<<<<<<<< + * cdef DTYPE_t iw, ih, box_area + * cdef DTYPE_t ua + */ + __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_np); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 29, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_1); + __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_zeros); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 29, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_2); + __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0; + __pyx_t_1 = __Pyx_PyInt_From_unsigned_int(__pyx_v_N); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 29, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_1); + __pyx_t_3 = __Pyx_PyInt_From_unsigned_int(__pyx_v_K); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 29, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_3); + __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 29, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_4); + __Pyx_GIVEREF(__pyx_t_1); + PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_1); + __Pyx_GIVEREF(__pyx_t_3); + PyTuple_SET_ITEM(__pyx_t_4, 1, __pyx_t_3); + __pyx_t_1 = 0; + __pyx_t_3 = 0; + __pyx_t_3 = PyTuple_New(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 29, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_3); + __Pyx_GIVEREF(__pyx_t_4); + PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_4); + __pyx_t_4 = 0; + __pyx_t_4 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 29, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_4); + __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_DTYPE); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 29, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_1); + if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_dtype, __pyx_t_1) < 0) __PYX_ERR(0, 29, __pyx_L1_error) + __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0; + __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_3, __pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 29, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_1); + __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0; + __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0; + __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0; + if (!(likely(((__pyx_t_1) == Py_None) || likely(__Pyx_TypeTest(__pyx_t_1, __pyx_ptype_5numpy_ndarray))))) __PYX_ERR(0, 29, __pyx_L1_error) + __pyx_t_5 = ((PyArrayObject *)__pyx_t_1); + { + __Pyx_BufFmt_StackElem __pyx_stack[1]; + if (unlikely(__Pyx_GetBufferAndValidate(&__pyx_pybuffernd_overlaps.rcbuffer->pybuffer, (PyObject*)__pyx_t_5, &__Pyx_TypeInfo_nn___pyx_t_4bbox_DTYPE_t, PyBUF_FORMAT| PyBUF_STRIDES| PyBUF_WRITABLE, 2, 0, __pyx_stack) == -1)) { + __pyx_v_overlaps = ((PyArrayObject *)Py_None); __Pyx_INCREF(Py_None); __pyx_pybuffernd_overlaps.rcbuffer->pybuffer.buf = NULL; + __PYX_ERR(0, 29, __pyx_L1_error) + } else {__pyx_pybuffernd_overlaps.diminfo[0].strides = __pyx_pybuffernd_overlaps.rcbuffer->pybuffer.strides[0]; __pyx_pybuffernd_overlaps.diminfo[0].shape = __pyx_pybuffernd_overlaps.rcbuffer->pybuffer.shape[0]; __pyx_pybuffernd_overlaps.diminfo[1].strides = __pyx_pybuffernd_overlaps.rcbuffer->pybuffer.strides[1]; __pyx_pybuffernd_overlaps.diminfo[1].shape = __pyx_pybuffernd_overlaps.rcbuffer->pybuffer.shape[1]; + } + } + __pyx_t_5 = 0; + __pyx_v_overlaps = ((PyArrayObject *)__pyx_t_1); + __pyx_t_1 = 0; + + /* "box_overlaps.pyx":33 + * cdef DTYPE_t ua + * cdef unsigned int k, n + * for k in range(K): # <<<<<<<<<<<<<< + * box_area = ( + * (query_boxes[k, 2] - query_boxes[k, 0] + 1) * + */ + __pyx_t_6 = __pyx_v_K; + __pyx_t_7 = __pyx_t_6; + for (__pyx_t_8 = 0; __pyx_t_8 < __pyx_t_7; __pyx_t_8+=1) { + __pyx_v_k = __pyx_t_8; + + /* "box_overlaps.pyx":35 + * for k in range(K): + * box_area = ( + * (query_boxes[k, 2] - query_boxes[k, 0] + 1) * # <<<<<<<<<<<<<< + * (query_boxes[k, 3] - query_boxes[k, 1] + 1) + * ) + */ + __pyx_t_9 = __pyx_v_k; + __pyx_t_10 = 2; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_9 >= (size_t)__pyx_pybuffernd_query_boxes.diminfo[0].shape)) __pyx_t_11 = 0; + if (__pyx_t_10 < 0) { + __pyx_t_10 += __pyx_pybuffernd_query_boxes.diminfo[1].shape; + if (unlikely(__pyx_t_10 < 0)) __pyx_t_11 = 1; + } else if (unlikely(__pyx_t_10 >= __pyx_pybuffernd_query_boxes.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 35, __pyx_L1_error) + } + __pyx_t_12 = __pyx_v_k; + __pyx_t_13 = 0; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_12 >= (size_t)__pyx_pybuffernd_query_boxes.diminfo[0].shape)) __pyx_t_11 = 0; + if (__pyx_t_13 < 0) { + __pyx_t_13 += __pyx_pybuffernd_query_boxes.diminfo[1].shape; + if (unlikely(__pyx_t_13 < 0)) __pyx_t_11 = 1; + } else if (unlikely(__pyx_t_13 >= __pyx_pybuffernd_query_boxes.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 35, __pyx_L1_error) + } + + /* "box_overlaps.pyx":36 + * box_area = ( + * (query_boxes[k, 2] - query_boxes[k, 0] + 1) * + * (query_boxes[k, 3] - query_boxes[k, 1] + 1) # <<<<<<<<<<<<<< + * ) + * for n in range(N): + */ + __pyx_t_14 = __pyx_v_k; + __pyx_t_15 = 3; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_14 >= (size_t)__pyx_pybuffernd_query_boxes.diminfo[0].shape)) __pyx_t_11 = 0; + if (__pyx_t_15 < 0) { + __pyx_t_15 += __pyx_pybuffernd_query_boxes.diminfo[1].shape; + if (unlikely(__pyx_t_15 < 0)) __pyx_t_11 = 1; + } else if (unlikely(__pyx_t_15 >= __pyx_pybuffernd_query_boxes.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 36, __pyx_L1_error) + } + __pyx_t_16 = __pyx_v_k; + __pyx_t_17 = 1; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_16 >= (size_t)__pyx_pybuffernd_query_boxes.diminfo[0].shape)) __pyx_t_11 = 0; + if (__pyx_t_17 < 0) { + __pyx_t_17 += __pyx_pybuffernd_query_boxes.diminfo[1].shape; + if (unlikely(__pyx_t_17 < 0)) __pyx_t_11 = 1; + } else if (unlikely(__pyx_t_17 >= __pyx_pybuffernd_query_boxes.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 36, __pyx_L1_error) + } + + /* "box_overlaps.pyx":35 + * for k in range(K): + * box_area = ( + * (query_boxes[k, 2] - query_boxes[k, 0] + 1) * # <<<<<<<<<<<<<< + * (query_boxes[k, 3] - query_boxes[k, 1] + 1) + * ) + */ + __pyx_v_box_area = ((((*__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_query_boxes.rcbuffer->pybuffer.buf, __pyx_t_9, __pyx_pybuffernd_query_boxes.diminfo[0].strides, __pyx_t_10, __pyx_pybuffernd_query_boxes.diminfo[1].strides)) - (*__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_query_boxes.rcbuffer->pybuffer.buf, __pyx_t_12, __pyx_pybuffernd_query_boxes.diminfo[0].strides, __pyx_t_13, __pyx_pybuffernd_query_boxes.diminfo[1].strides))) + 1.0) * (((*__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_query_boxes.rcbuffer->pybuffer.buf, __pyx_t_14, __pyx_pybuffernd_query_boxes.diminfo[0].strides, __pyx_t_15, __pyx_pybuffernd_query_boxes.diminfo[1].strides)) - (*__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_query_boxes.rcbuffer->pybuffer.buf, __pyx_t_16, __pyx_pybuffernd_query_boxes.diminfo[0].strides, __pyx_t_17, __pyx_pybuffernd_query_boxes.diminfo[1].strides))) + 1.0)); + + /* "box_overlaps.pyx":38 + * (query_boxes[k, 3] - query_boxes[k, 1] + 1) + * ) + * for n in range(N): # <<<<<<<<<<<<<< + * iw = ( + * min(boxes[n, 2], query_boxes[k, 2]) - + */ + __pyx_t_18 = __pyx_v_N; + __pyx_t_19 = __pyx_t_18; + for (__pyx_t_20 = 0; __pyx_t_20 < __pyx_t_19; __pyx_t_20+=1) { + __pyx_v_n = __pyx_t_20; + + /* "box_overlaps.pyx":40 + * for n in range(N): + * iw = ( + * min(boxes[n, 2], query_boxes[k, 2]) - # <<<<<<<<<<<<<< + * max(boxes[n, 0], query_boxes[k, 0]) + 1 + * ) + */ + __pyx_t_16 = __pyx_v_k; + __pyx_t_17 = 2; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_16 >= (size_t)__pyx_pybuffernd_query_boxes.diminfo[0].shape)) __pyx_t_11 = 0; + if (__pyx_t_17 < 0) { + __pyx_t_17 += __pyx_pybuffernd_query_boxes.diminfo[1].shape; + if (unlikely(__pyx_t_17 < 0)) __pyx_t_11 = 1; + } else if (unlikely(__pyx_t_17 >= __pyx_pybuffernd_query_boxes.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 40, __pyx_L1_error) + } + __pyx_t_21 = (*__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_query_boxes.rcbuffer->pybuffer.buf, __pyx_t_16, __pyx_pybuffernd_query_boxes.diminfo[0].strides, __pyx_t_17, __pyx_pybuffernd_query_boxes.diminfo[1].strides)); + __pyx_t_16 = __pyx_v_n; + __pyx_t_17 = 2; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_16 >= (size_t)__pyx_pybuffernd_boxes.diminfo[0].shape)) __pyx_t_11 = 0; + if (__pyx_t_17 < 0) { + __pyx_t_17 += __pyx_pybuffernd_boxes.diminfo[1].shape; + if (unlikely(__pyx_t_17 < 0)) __pyx_t_11 = 1; + } else if (unlikely(__pyx_t_17 >= __pyx_pybuffernd_boxes.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 40, __pyx_L1_error) + } + __pyx_t_22 = (*__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_boxes.rcbuffer->pybuffer.buf, __pyx_t_16, __pyx_pybuffernd_boxes.diminfo[0].strides, __pyx_t_17, __pyx_pybuffernd_boxes.diminfo[1].strides)); + if (((__pyx_t_21 < __pyx_t_22) != 0)) { + __pyx_t_23 = __pyx_t_21; + } else { + __pyx_t_23 = __pyx_t_22; + } + + /* "box_overlaps.pyx":41 + * iw = ( + * min(boxes[n, 2], query_boxes[k, 2]) - + * max(boxes[n, 0], query_boxes[k, 0]) + 1 # <<<<<<<<<<<<<< + * ) + * if iw > 0: + */ + __pyx_t_16 = __pyx_v_k; + __pyx_t_17 = 0; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_16 >= (size_t)__pyx_pybuffernd_query_boxes.diminfo[0].shape)) __pyx_t_11 = 0; + if (__pyx_t_17 < 0) { + __pyx_t_17 += __pyx_pybuffernd_query_boxes.diminfo[1].shape; + if (unlikely(__pyx_t_17 < 0)) __pyx_t_11 = 1; + } else if (unlikely(__pyx_t_17 >= __pyx_pybuffernd_query_boxes.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 41, __pyx_L1_error) + } + __pyx_t_21 = (*__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_query_boxes.rcbuffer->pybuffer.buf, __pyx_t_16, __pyx_pybuffernd_query_boxes.diminfo[0].strides, __pyx_t_17, __pyx_pybuffernd_query_boxes.diminfo[1].strides)); + __pyx_t_16 = __pyx_v_n; + __pyx_t_17 = 0; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_16 >= (size_t)__pyx_pybuffernd_boxes.diminfo[0].shape)) __pyx_t_11 = 0; + if (__pyx_t_17 < 0) { + __pyx_t_17 += __pyx_pybuffernd_boxes.diminfo[1].shape; + if (unlikely(__pyx_t_17 < 0)) __pyx_t_11 = 1; + } else if (unlikely(__pyx_t_17 >= __pyx_pybuffernd_boxes.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 41, __pyx_L1_error) + } + __pyx_t_22 = (*__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_boxes.rcbuffer->pybuffer.buf, __pyx_t_16, __pyx_pybuffernd_boxes.diminfo[0].strides, __pyx_t_17, __pyx_pybuffernd_boxes.diminfo[1].strides)); + if (((__pyx_t_21 > __pyx_t_22) != 0)) { + __pyx_t_24 = __pyx_t_21; + } else { + __pyx_t_24 = __pyx_t_22; + } + + /* "box_overlaps.pyx":40 + * for n in range(N): + * iw = ( + * min(boxes[n, 2], query_boxes[k, 2]) - # <<<<<<<<<<<<<< + * max(boxes[n, 0], query_boxes[k, 0]) + 1 + * ) + */ + __pyx_v_iw = ((__pyx_t_23 - __pyx_t_24) + 1.0); + + /* "box_overlaps.pyx":43 + * max(boxes[n, 0], query_boxes[k, 0]) + 1 + * ) + * if iw > 0: # <<<<<<<<<<<<<< + * ih = ( + * min(boxes[n, 3], query_boxes[k, 3]) - + */ + __pyx_t_25 = ((__pyx_v_iw > 0.0) != 0); + if (__pyx_t_25) { + + /* "box_overlaps.pyx":45 + * if iw > 0: + * ih = ( + * min(boxes[n, 3], query_boxes[k, 3]) - # <<<<<<<<<<<<<< + * max(boxes[n, 1], query_boxes[k, 1]) + 1 + * ) + */ + __pyx_t_16 = __pyx_v_k; + __pyx_t_17 = 3; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_16 >= (size_t)__pyx_pybuffernd_query_boxes.diminfo[0].shape)) __pyx_t_11 = 0; + if (__pyx_t_17 < 0) { + __pyx_t_17 += __pyx_pybuffernd_query_boxes.diminfo[1].shape; + if (unlikely(__pyx_t_17 < 0)) __pyx_t_11 = 1; + } else if (unlikely(__pyx_t_17 >= __pyx_pybuffernd_query_boxes.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 45, __pyx_L1_error) + } + __pyx_t_24 = (*__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_query_boxes.rcbuffer->pybuffer.buf, __pyx_t_16, __pyx_pybuffernd_query_boxes.diminfo[0].strides, __pyx_t_17, __pyx_pybuffernd_query_boxes.diminfo[1].strides)); + __pyx_t_16 = __pyx_v_n; + __pyx_t_17 = 3; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_16 >= (size_t)__pyx_pybuffernd_boxes.diminfo[0].shape)) __pyx_t_11 = 0; + if (__pyx_t_17 < 0) { + __pyx_t_17 += __pyx_pybuffernd_boxes.diminfo[1].shape; + if (unlikely(__pyx_t_17 < 0)) __pyx_t_11 = 1; + } else if (unlikely(__pyx_t_17 >= __pyx_pybuffernd_boxes.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 45, __pyx_L1_error) + } + __pyx_t_23 = (*__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_boxes.rcbuffer->pybuffer.buf, __pyx_t_16, __pyx_pybuffernd_boxes.diminfo[0].strides, __pyx_t_17, __pyx_pybuffernd_boxes.diminfo[1].strides)); + if (((__pyx_t_24 < __pyx_t_23) != 0)) { + __pyx_t_21 = __pyx_t_24; + } else { + __pyx_t_21 = __pyx_t_23; + } + + /* "box_overlaps.pyx":46 + * ih = ( + * min(boxes[n, 3], query_boxes[k, 3]) - + * max(boxes[n, 1], query_boxes[k, 1]) + 1 # <<<<<<<<<<<<<< + * ) + * if ih > 0: + */ + __pyx_t_16 = __pyx_v_k; + __pyx_t_17 = 1; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_16 >= (size_t)__pyx_pybuffernd_query_boxes.diminfo[0].shape)) __pyx_t_11 = 0; + if (__pyx_t_17 < 0) { + __pyx_t_17 += __pyx_pybuffernd_query_boxes.diminfo[1].shape; + if (unlikely(__pyx_t_17 < 0)) __pyx_t_11 = 1; + } else if (unlikely(__pyx_t_17 >= __pyx_pybuffernd_query_boxes.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 46, __pyx_L1_error) + } + __pyx_t_24 = (*__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_query_boxes.rcbuffer->pybuffer.buf, __pyx_t_16, __pyx_pybuffernd_query_boxes.diminfo[0].strides, __pyx_t_17, __pyx_pybuffernd_query_boxes.diminfo[1].strides)); + __pyx_t_16 = __pyx_v_n; + __pyx_t_17 = 1; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_16 >= (size_t)__pyx_pybuffernd_boxes.diminfo[0].shape)) __pyx_t_11 = 0; + if (__pyx_t_17 < 0) { + __pyx_t_17 += __pyx_pybuffernd_boxes.diminfo[1].shape; + if (unlikely(__pyx_t_17 < 0)) __pyx_t_11 = 1; + } else if (unlikely(__pyx_t_17 >= __pyx_pybuffernd_boxes.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 46, __pyx_L1_error) + } + __pyx_t_23 = (*__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_boxes.rcbuffer->pybuffer.buf, __pyx_t_16, __pyx_pybuffernd_boxes.diminfo[0].strides, __pyx_t_17, __pyx_pybuffernd_boxes.diminfo[1].strides)); + if (((__pyx_t_24 > __pyx_t_23) != 0)) { + __pyx_t_22 = __pyx_t_24; + } else { + __pyx_t_22 = __pyx_t_23; + } + + /* "box_overlaps.pyx":45 + * if iw > 0: + * ih = ( + * min(boxes[n, 3], query_boxes[k, 3]) - # <<<<<<<<<<<<<< + * max(boxes[n, 1], query_boxes[k, 1]) + 1 + * ) + */ + __pyx_v_ih = ((__pyx_t_21 - __pyx_t_22) + 1.0); + + /* "box_overlaps.pyx":48 + * max(boxes[n, 1], query_boxes[k, 1]) + 1 + * ) + * if ih > 0: # <<<<<<<<<<<<<< + * ua = float( + * (boxes[n, 2] - boxes[n, 0] + 1) * + */ + __pyx_t_25 = ((__pyx_v_ih > 0.0) != 0); + if (__pyx_t_25) { + + /* "box_overlaps.pyx":50 + * if ih > 0: + * ua = float( + * (boxes[n, 2] - boxes[n, 0] + 1) * # <<<<<<<<<<<<<< + * (boxes[n, 3] - boxes[n, 1] + 1) + + * box_area - iw * ih + */ + __pyx_t_16 = __pyx_v_n; + __pyx_t_17 = 2; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_16 >= (size_t)__pyx_pybuffernd_boxes.diminfo[0].shape)) __pyx_t_11 = 0; + if (__pyx_t_17 < 0) { + __pyx_t_17 += __pyx_pybuffernd_boxes.diminfo[1].shape; + if (unlikely(__pyx_t_17 < 0)) __pyx_t_11 = 1; + } else if (unlikely(__pyx_t_17 >= __pyx_pybuffernd_boxes.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 50, __pyx_L1_error) + } + __pyx_t_14 = __pyx_v_n; + __pyx_t_15 = 0; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_14 >= (size_t)__pyx_pybuffernd_boxes.diminfo[0].shape)) __pyx_t_11 = 0; + if (__pyx_t_15 < 0) { + __pyx_t_15 += __pyx_pybuffernd_boxes.diminfo[1].shape; + if (unlikely(__pyx_t_15 < 0)) __pyx_t_11 = 1; + } else if (unlikely(__pyx_t_15 >= __pyx_pybuffernd_boxes.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 50, __pyx_L1_error) + } + + /* "box_overlaps.pyx":51 + * ua = float( + * (boxes[n, 2] - boxes[n, 0] + 1) * + * (boxes[n, 3] - boxes[n, 1] + 1) + # <<<<<<<<<<<<<< + * box_area - iw * ih + * ) + */ + __pyx_t_12 = __pyx_v_n; + __pyx_t_13 = 3; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_12 >= (size_t)__pyx_pybuffernd_boxes.diminfo[0].shape)) __pyx_t_11 = 0; + if (__pyx_t_13 < 0) { + __pyx_t_13 += __pyx_pybuffernd_boxes.diminfo[1].shape; + if (unlikely(__pyx_t_13 < 0)) __pyx_t_11 = 1; + } else if (unlikely(__pyx_t_13 >= __pyx_pybuffernd_boxes.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 51, __pyx_L1_error) + } + __pyx_t_9 = __pyx_v_n; + __pyx_t_10 = 1; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_9 >= (size_t)__pyx_pybuffernd_boxes.diminfo[0].shape)) __pyx_t_11 = 0; + if (__pyx_t_10 < 0) { + __pyx_t_10 += __pyx_pybuffernd_boxes.diminfo[1].shape; + if (unlikely(__pyx_t_10 < 0)) __pyx_t_11 = 1; + } else if (unlikely(__pyx_t_10 >= __pyx_pybuffernd_boxes.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 51, __pyx_L1_error) + } + + /* "box_overlaps.pyx":49 + * ) + * if ih > 0: + * ua = float( # <<<<<<<<<<<<<< + * (boxes[n, 2] - boxes[n, 0] + 1) * + * (boxes[n, 3] - boxes[n, 1] + 1) + + */ + __pyx_v_ua = ((double)((((((*__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_boxes.rcbuffer->pybuffer.buf, __pyx_t_16, __pyx_pybuffernd_boxes.diminfo[0].strides, __pyx_t_17, __pyx_pybuffernd_boxes.diminfo[1].strides)) - (*__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_boxes.rcbuffer->pybuffer.buf, __pyx_t_14, __pyx_pybuffernd_boxes.diminfo[0].strides, __pyx_t_15, __pyx_pybuffernd_boxes.diminfo[1].strides))) + 1.0) * (((*__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_boxes.rcbuffer->pybuffer.buf, __pyx_t_12, __pyx_pybuffernd_boxes.diminfo[0].strides, __pyx_t_13, __pyx_pybuffernd_boxes.diminfo[1].strides)) - (*__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_boxes.rcbuffer->pybuffer.buf, __pyx_t_9, __pyx_pybuffernd_boxes.diminfo[0].strides, __pyx_t_10, __pyx_pybuffernd_boxes.diminfo[1].strides))) + 1.0)) + __pyx_v_box_area) - (__pyx_v_iw * __pyx_v_ih))); + + /* "box_overlaps.pyx":54 + * box_area - iw * ih + * ) + * overlaps[n, k] = iw * ih / ua # <<<<<<<<<<<<<< + * return overlaps + */ + __pyx_t_22 = (__pyx_v_iw * __pyx_v_ih); + if (unlikely(__pyx_v_ua == 0)) { + PyErr_SetString(PyExc_ZeroDivisionError, "float division"); + __PYX_ERR(0, 54, __pyx_L1_error) + } + __pyx_t_9 = __pyx_v_n; + __pyx_t_12 = __pyx_v_k; + __pyx_t_11 = -1; + if (unlikely(__pyx_t_9 >= (size_t)__pyx_pybuffernd_overlaps.diminfo[0].shape)) __pyx_t_11 = 0; + if (unlikely(__pyx_t_12 >= (size_t)__pyx_pybuffernd_overlaps.diminfo[1].shape)) __pyx_t_11 = 1; + if (unlikely(__pyx_t_11 != -1)) { + __Pyx_RaiseBufferIndexError(__pyx_t_11); + __PYX_ERR(0, 54, __pyx_L1_error) + } + *__Pyx_BufPtrStrided2d(__pyx_t_4bbox_DTYPE_t *, __pyx_pybuffernd_overlaps.rcbuffer->pybuffer.buf, __pyx_t_9, __pyx_pybuffernd_overlaps.diminfo[0].strides, __pyx_t_12, __pyx_pybuffernd_overlaps.diminfo[1].strides) = (__pyx_t_22 / __pyx_v_ua); + + /* "box_overlaps.pyx":48 + * max(boxes[n, 1], query_boxes[k, 1]) + 1 + * ) + * if ih > 0: # <<<<<<<<<<<<<< + * ua = float( + * (boxes[n, 2] - boxes[n, 0] + 1) * + */ + } + + /* "box_overlaps.pyx":43 + * max(boxes[n, 0], query_boxes[k, 0]) + 1 + * ) + * if iw > 0: # <<<<<<<<<<<<<< + * ih = ( + * min(boxes[n, 3], query_boxes[k, 3]) - + */ + } + } + } + + /* "box_overlaps.pyx":55 + * ) + * overlaps[n, k] = iw * ih / ua + * return overlaps # <<<<<<<<<<<<<< + */ + __Pyx_XDECREF(__pyx_r); + __Pyx_INCREF(((PyObject *)__pyx_v_overlaps)); + __pyx_r = ((PyObject *)__pyx_v_overlaps); + goto __pyx_L0; + + /* "box_overlaps.pyx":15 + * ctypedef np.float_t DTYPE_t + * + * def bbox_overlaps( # <<<<<<<<<<<<<< + * np.ndarray[DTYPE_t, ndim=2] boxes, + * np.ndarray[DTYPE_t, ndim=2] query_boxes): + */ + + /* function exit code */ + __pyx_L1_error:; + __Pyx_XDECREF(__pyx_t_1); + __Pyx_XDECREF(__pyx_t_2); + __Pyx_XDECREF(__pyx_t_3); + __Pyx_XDECREF(__pyx_t_4); + { PyObject *__pyx_type, *__pyx_value, *__pyx_tb; + __Pyx_PyThreadState_declare + __Pyx_PyThreadState_assign + __Pyx_ErrFetch(&__pyx_type, &__pyx_value, &__pyx_tb); + __Pyx_SafeReleaseBuffer(&__pyx_pybuffernd_boxes.rcbuffer->pybuffer); + __Pyx_SafeReleaseBuffer(&__pyx_pybuffernd_overlaps.rcbuffer->pybuffer); + __Pyx_SafeReleaseBuffer(&__pyx_pybuffernd_query_boxes.rcbuffer->pybuffer); + __Pyx_ErrRestore(__pyx_type, __pyx_value, __pyx_tb);} + __Pyx_AddTraceback("bbox.bbox_overlaps", __pyx_clineno, __pyx_lineno, __pyx_filename); + __pyx_r = NULL; + goto __pyx_L2; + __pyx_L0:; + __Pyx_SafeReleaseBuffer(&__pyx_pybuffernd_boxes.rcbuffer->pybuffer); + __Pyx_SafeReleaseBuffer(&__pyx_pybuffernd_overlaps.rcbuffer->pybuffer); + __Pyx_SafeReleaseBuffer(&__pyx_pybuffernd_query_boxes.rcbuffer->pybuffer); + __pyx_L2:; + __Pyx_XDECREF((PyObject *)__pyx_v_overlaps); + __Pyx_XGIVEREF(__pyx_r); + __Pyx_RefNannyFinishContext(); + return __pyx_r; +} + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":734 + * ctypedef npy_cdouble complex_t + * + * cdef inline object PyArray_MultiIterNew1(a): # <<<<<<<<<<<<<< + * return PyArray_MultiIterNew(1, a) + * + */ + +static CYTHON_INLINE PyObject *__pyx_f_5numpy_PyArray_MultiIterNew1(PyObject *__pyx_v_a) { + PyObject *__pyx_r = NULL; + __Pyx_RefNannyDeclarations + PyObject *__pyx_t_1 = NULL; + int __pyx_lineno = 0; + const char *__pyx_filename = NULL; + int __pyx_clineno = 0; + __Pyx_RefNannySetupContext("PyArray_MultiIterNew1", 0); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":735 + * + * cdef inline object PyArray_MultiIterNew1(a): + * return PyArray_MultiIterNew(1, a) # <<<<<<<<<<<<<< + * + * cdef inline object PyArray_MultiIterNew2(a, b): + */ + __Pyx_XDECREF(__pyx_r); + __pyx_t_1 = PyArray_MultiIterNew(1, ((void *)__pyx_v_a)); if (unlikely(!__pyx_t_1)) __PYX_ERR(1, 735, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_1); + __pyx_r = __pyx_t_1; + __pyx_t_1 = 0; + goto __pyx_L0; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":734 + * ctypedef npy_cdouble complex_t + * + * cdef inline object PyArray_MultiIterNew1(a): # <<<<<<<<<<<<<< + * return PyArray_MultiIterNew(1, a) + * + */ + + /* function exit code */ + __pyx_L1_error:; + __Pyx_XDECREF(__pyx_t_1); + __Pyx_AddTraceback("numpy.PyArray_MultiIterNew1", __pyx_clineno, __pyx_lineno, __pyx_filename); + __pyx_r = 0; + __pyx_L0:; + __Pyx_XGIVEREF(__pyx_r); + __Pyx_RefNannyFinishContext(); + return __pyx_r; +} + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":737 + * return PyArray_MultiIterNew(1, a) + * + * cdef inline object PyArray_MultiIterNew2(a, b): # <<<<<<<<<<<<<< + * return PyArray_MultiIterNew(2, a, b) + * + */ + +static CYTHON_INLINE PyObject *__pyx_f_5numpy_PyArray_MultiIterNew2(PyObject *__pyx_v_a, PyObject *__pyx_v_b) { + PyObject *__pyx_r = NULL; + __Pyx_RefNannyDeclarations + PyObject *__pyx_t_1 = NULL; + int __pyx_lineno = 0; + const char *__pyx_filename = NULL; + int __pyx_clineno = 0; + __Pyx_RefNannySetupContext("PyArray_MultiIterNew2", 0); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":738 + * + * cdef inline object PyArray_MultiIterNew2(a, b): + * return PyArray_MultiIterNew(2, a, b) # <<<<<<<<<<<<<< + * + * cdef inline object PyArray_MultiIterNew3(a, b, c): + */ + __Pyx_XDECREF(__pyx_r); + __pyx_t_1 = PyArray_MultiIterNew(2, ((void *)__pyx_v_a), ((void *)__pyx_v_b)); if (unlikely(!__pyx_t_1)) __PYX_ERR(1, 738, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_1); + __pyx_r = __pyx_t_1; + __pyx_t_1 = 0; + goto __pyx_L0; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":737 + * return PyArray_MultiIterNew(1, a) + * + * cdef inline object PyArray_MultiIterNew2(a, b): # <<<<<<<<<<<<<< + * return PyArray_MultiIterNew(2, a, b) + * + */ + + /* function exit code */ + __pyx_L1_error:; + __Pyx_XDECREF(__pyx_t_1); + __Pyx_AddTraceback("numpy.PyArray_MultiIterNew2", __pyx_clineno, __pyx_lineno, __pyx_filename); + __pyx_r = 0; + __pyx_L0:; + __Pyx_XGIVEREF(__pyx_r); + __Pyx_RefNannyFinishContext(); + return __pyx_r; +} + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":740 + * return PyArray_MultiIterNew(2, a, b) + * + * cdef inline object PyArray_MultiIterNew3(a, b, c): # <<<<<<<<<<<<<< + * return PyArray_MultiIterNew(3, a, b, c) + * + */ + +static CYTHON_INLINE PyObject *__pyx_f_5numpy_PyArray_MultiIterNew3(PyObject *__pyx_v_a, PyObject *__pyx_v_b, PyObject *__pyx_v_c) { + PyObject *__pyx_r = NULL; + __Pyx_RefNannyDeclarations + PyObject *__pyx_t_1 = NULL; + int __pyx_lineno = 0; + const char *__pyx_filename = NULL; + int __pyx_clineno = 0; + __Pyx_RefNannySetupContext("PyArray_MultiIterNew3", 0); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":741 + * + * cdef inline object PyArray_MultiIterNew3(a, b, c): + * return PyArray_MultiIterNew(3, a, b, c) # <<<<<<<<<<<<<< + * + * cdef inline object PyArray_MultiIterNew4(a, b, c, d): + */ + __Pyx_XDECREF(__pyx_r); + __pyx_t_1 = PyArray_MultiIterNew(3, ((void *)__pyx_v_a), ((void *)__pyx_v_b), ((void *)__pyx_v_c)); if (unlikely(!__pyx_t_1)) __PYX_ERR(1, 741, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_1); + __pyx_r = __pyx_t_1; + __pyx_t_1 = 0; + goto __pyx_L0; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":740 + * return PyArray_MultiIterNew(2, a, b) + * + * cdef inline object PyArray_MultiIterNew3(a, b, c): # <<<<<<<<<<<<<< + * return PyArray_MultiIterNew(3, a, b, c) + * + */ + + /* function exit code */ + __pyx_L1_error:; + __Pyx_XDECREF(__pyx_t_1); + __Pyx_AddTraceback("numpy.PyArray_MultiIterNew3", __pyx_clineno, __pyx_lineno, __pyx_filename); + __pyx_r = 0; + __pyx_L0:; + __Pyx_XGIVEREF(__pyx_r); + __Pyx_RefNannyFinishContext(); + return __pyx_r; +} + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":743 + * return PyArray_MultiIterNew(3, a, b, c) + * + * cdef inline object PyArray_MultiIterNew4(a, b, c, d): # <<<<<<<<<<<<<< + * return PyArray_MultiIterNew(4, a, b, c, d) + * + */ + +static CYTHON_INLINE PyObject *__pyx_f_5numpy_PyArray_MultiIterNew4(PyObject *__pyx_v_a, PyObject *__pyx_v_b, PyObject *__pyx_v_c, PyObject *__pyx_v_d) { + PyObject *__pyx_r = NULL; + __Pyx_RefNannyDeclarations + PyObject *__pyx_t_1 = NULL; + int __pyx_lineno = 0; + const char *__pyx_filename = NULL; + int __pyx_clineno = 0; + __Pyx_RefNannySetupContext("PyArray_MultiIterNew4", 0); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":744 + * + * cdef inline object PyArray_MultiIterNew4(a, b, c, d): + * return PyArray_MultiIterNew(4, a, b, c, d) # <<<<<<<<<<<<<< + * + * cdef inline object PyArray_MultiIterNew5(a, b, c, d, e): + */ + __Pyx_XDECREF(__pyx_r); + __pyx_t_1 = PyArray_MultiIterNew(4, ((void *)__pyx_v_a), ((void *)__pyx_v_b), ((void *)__pyx_v_c), ((void *)__pyx_v_d)); if (unlikely(!__pyx_t_1)) __PYX_ERR(1, 744, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_1); + __pyx_r = __pyx_t_1; + __pyx_t_1 = 0; + goto __pyx_L0; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":743 + * return PyArray_MultiIterNew(3, a, b, c) + * + * cdef inline object PyArray_MultiIterNew4(a, b, c, d): # <<<<<<<<<<<<<< + * return PyArray_MultiIterNew(4, a, b, c, d) + * + */ + + /* function exit code */ + __pyx_L1_error:; + __Pyx_XDECREF(__pyx_t_1); + __Pyx_AddTraceback("numpy.PyArray_MultiIterNew4", __pyx_clineno, __pyx_lineno, __pyx_filename); + __pyx_r = 0; + __pyx_L0:; + __Pyx_XGIVEREF(__pyx_r); + __Pyx_RefNannyFinishContext(); + return __pyx_r; +} + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":746 + * return PyArray_MultiIterNew(4, a, b, c, d) + * + * cdef inline object PyArray_MultiIterNew5(a, b, c, d, e): # <<<<<<<<<<<<<< + * return PyArray_MultiIterNew(5, a, b, c, d, e) + * + */ + +static CYTHON_INLINE PyObject *__pyx_f_5numpy_PyArray_MultiIterNew5(PyObject *__pyx_v_a, PyObject *__pyx_v_b, PyObject *__pyx_v_c, PyObject *__pyx_v_d, PyObject *__pyx_v_e) { + PyObject *__pyx_r = NULL; + __Pyx_RefNannyDeclarations + PyObject *__pyx_t_1 = NULL; + int __pyx_lineno = 0; + const char *__pyx_filename = NULL; + int __pyx_clineno = 0; + __Pyx_RefNannySetupContext("PyArray_MultiIterNew5", 0); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":747 + * + * cdef inline object PyArray_MultiIterNew5(a, b, c, d, e): + * return PyArray_MultiIterNew(5, a, b, c, d, e) # <<<<<<<<<<<<<< + * + * cdef inline tuple PyDataType_SHAPE(dtype d): + */ + __Pyx_XDECREF(__pyx_r); + __pyx_t_1 = PyArray_MultiIterNew(5, ((void *)__pyx_v_a), ((void *)__pyx_v_b), ((void *)__pyx_v_c), ((void *)__pyx_v_d), ((void *)__pyx_v_e)); if (unlikely(!__pyx_t_1)) __PYX_ERR(1, 747, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_1); + __pyx_r = __pyx_t_1; + __pyx_t_1 = 0; + goto __pyx_L0; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":746 + * return PyArray_MultiIterNew(4, a, b, c, d) + * + * cdef inline object PyArray_MultiIterNew5(a, b, c, d, e): # <<<<<<<<<<<<<< + * return PyArray_MultiIterNew(5, a, b, c, d, e) + * + */ + + /* function exit code */ + __pyx_L1_error:; + __Pyx_XDECREF(__pyx_t_1); + __Pyx_AddTraceback("numpy.PyArray_MultiIterNew5", __pyx_clineno, __pyx_lineno, __pyx_filename); + __pyx_r = 0; + __pyx_L0:; + __Pyx_XGIVEREF(__pyx_r); + __Pyx_RefNannyFinishContext(); + return __pyx_r; +} + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":749 + * return PyArray_MultiIterNew(5, a, b, c, d, e) + * + * cdef inline tuple PyDataType_SHAPE(dtype d): # <<<<<<<<<<<<<< + * if PyDataType_HASSUBARRAY(d): + * return d.subarray.shape + */ + +static CYTHON_INLINE PyObject *__pyx_f_5numpy_PyDataType_SHAPE(PyArray_Descr *__pyx_v_d) { + PyObject *__pyx_r = NULL; + __Pyx_RefNannyDeclarations + int __pyx_t_1; + __Pyx_RefNannySetupContext("PyDataType_SHAPE", 0); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":750 + * + * cdef inline tuple PyDataType_SHAPE(dtype d): + * if PyDataType_HASSUBARRAY(d): # <<<<<<<<<<<<<< + * return d.subarray.shape + * else: + */ + __pyx_t_1 = (PyDataType_HASSUBARRAY(__pyx_v_d) != 0); + if (__pyx_t_1) { + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":751 + * cdef inline tuple PyDataType_SHAPE(dtype d): + * if PyDataType_HASSUBARRAY(d): + * return d.subarray.shape # <<<<<<<<<<<<<< + * else: + * return () + */ + __Pyx_XDECREF(__pyx_r); + __Pyx_INCREF(((PyObject*)__pyx_v_d->subarray->shape)); + __pyx_r = ((PyObject*)__pyx_v_d->subarray->shape); + goto __pyx_L0; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":750 + * + * cdef inline tuple PyDataType_SHAPE(dtype d): + * if PyDataType_HASSUBARRAY(d): # <<<<<<<<<<<<<< + * return d.subarray.shape + * else: + */ + } + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":753 + * return d.subarray.shape + * else: + * return () # <<<<<<<<<<<<<< + * + * + */ + /*else*/ { + __Pyx_XDECREF(__pyx_r); + __Pyx_INCREF(__pyx_empty_tuple); + __pyx_r = __pyx_empty_tuple; + goto __pyx_L0; + } + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":749 + * return PyArray_MultiIterNew(5, a, b, c, d, e) + * + * cdef inline tuple PyDataType_SHAPE(dtype d): # <<<<<<<<<<<<<< + * if PyDataType_HASSUBARRAY(d): + * return d.subarray.shape + */ + + /* function exit code */ + __pyx_L0:; + __Pyx_XGIVEREF(__pyx_r); + __Pyx_RefNannyFinishContext(); + return __pyx_r; +} + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":928 + * int _import_umath() except -1 + * + * cdef inline void set_array_base(ndarray arr, object base): # <<<<<<<<<<<<<< + * Py_INCREF(base) # important to do this before stealing the reference below! + * PyArray_SetBaseObject(arr, base) + */ + +static CYTHON_INLINE void __pyx_f_5numpy_set_array_base(PyArrayObject *__pyx_v_arr, PyObject *__pyx_v_base) { + __Pyx_RefNannyDeclarations + __Pyx_RefNannySetupContext("set_array_base", 0); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":929 + * + * cdef inline void set_array_base(ndarray arr, object base): + * Py_INCREF(base) # important to do this before stealing the reference below! # <<<<<<<<<<<<<< + * PyArray_SetBaseObject(arr, base) + * + */ + Py_INCREF(__pyx_v_base); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":930 + * cdef inline void set_array_base(ndarray arr, object base): + * Py_INCREF(base) # important to do this before stealing the reference below! + * PyArray_SetBaseObject(arr, base) # <<<<<<<<<<<<<< + * + * cdef inline object get_array_base(ndarray arr): + */ + (void)(PyArray_SetBaseObject(__pyx_v_arr, __pyx_v_base)); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":928 + * int _import_umath() except -1 + * + * cdef inline void set_array_base(ndarray arr, object base): # <<<<<<<<<<<<<< + * Py_INCREF(base) # important to do this before stealing the reference below! + * PyArray_SetBaseObject(arr, base) + */ + + /* function exit code */ + __Pyx_RefNannyFinishContext(); +} + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":932 + * PyArray_SetBaseObject(arr, base) + * + * cdef inline object get_array_base(ndarray arr): # <<<<<<<<<<<<<< + * base = PyArray_BASE(arr) + * if base is NULL: + */ + +static CYTHON_INLINE PyObject *__pyx_f_5numpy_get_array_base(PyArrayObject *__pyx_v_arr) { + PyObject *__pyx_v_base; + PyObject *__pyx_r = NULL; + __Pyx_RefNannyDeclarations + int __pyx_t_1; + __Pyx_RefNannySetupContext("get_array_base", 0); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":933 + * + * cdef inline object get_array_base(ndarray arr): + * base = PyArray_BASE(arr) # <<<<<<<<<<<<<< + * if base is NULL: + * return None + */ + __pyx_v_base = PyArray_BASE(__pyx_v_arr); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":934 + * cdef inline object get_array_base(ndarray arr): + * base = PyArray_BASE(arr) + * if base is NULL: # <<<<<<<<<<<<<< + * return None + * return base + */ + __pyx_t_1 = ((__pyx_v_base == NULL) != 0); + if (__pyx_t_1) { + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":935 + * base = PyArray_BASE(arr) + * if base is NULL: + * return None # <<<<<<<<<<<<<< + * return base + * + */ + __Pyx_XDECREF(__pyx_r); + __pyx_r = Py_None; __Pyx_INCREF(Py_None); + goto __pyx_L0; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":934 + * cdef inline object get_array_base(ndarray arr): + * base = PyArray_BASE(arr) + * if base is NULL: # <<<<<<<<<<<<<< + * return None + * return base + */ + } + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":936 + * if base is NULL: + * return None + * return base # <<<<<<<<<<<<<< + * + * # Versions of the import_* functions which are more suitable for + */ + __Pyx_XDECREF(__pyx_r); + __Pyx_INCREF(((PyObject *)__pyx_v_base)); + __pyx_r = ((PyObject *)__pyx_v_base); + goto __pyx_L0; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":932 + * PyArray_SetBaseObject(arr, base) + * + * cdef inline object get_array_base(ndarray arr): # <<<<<<<<<<<<<< + * base = PyArray_BASE(arr) + * if base is NULL: + */ + + /* function exit code */ + __pyx_L0:; + __Pyx_XGIVEREF(__pyx_r); + __Pyx_RefNannyFinishContext(); + return __pyx_r; +} + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":940 + * # Versions of the import_* functions which are more suitable for + * # Cython code. + * cdef inline int import_array() except -1: # <<<<<<<<<<<<<< + * try: + * __pyx_import_array() + */ + +static CYTHON_INLINE int __pyx_f_5numpy_import_array(void) { + int __pyx_r; + __Pyx_RefNannyDeclarations + PyObject *__pyx_t_1 = NULL; + PyObject *__pyx_t_2 = NULL; + PyObject *__pyx_t_3 = NULL; + int __pyx_t_4; + PyObject *__pyx_t_5 = NULL; + PyObject *__pyx_t_6 = NULL; + PyObject *__pyx_t_7 = NULL; + PyObject *__pyx_t_8 = NULL; + int __pyx_lineno = 0; + const char *__pyx_filename = NULL; + int __pyx_clineno = 0; + __Pyx_RefNannySetupContext("import_array", 0); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":941 + * # Cython code. + * cdef inline int import_array() except -1: + * try: # <<<<<<<<<<<<<< + * __pyx_import_array() + * except Exception: + */ + { + __Pyx_PyThreadState_declare + __Pyx_PyThreadState_assign + __Pyx_ExceptionSave(&__pyx_t_1, &__pyx_t_2, &__pyx_t_3); + __Pyx_XGOTREF(__pyx_t_1); + __Pyx_XGOTREF(__pyx_t_2); + __Pyx_XGOTREF(__pyx_t_3); + /*try:*/ { + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":942 + * cdef inline int import_array() except -1: + * try: + * __pyx_import_array() # <<<<<<<<<<<<<< + * except Exception: + * raise ImportError("numpy.core.multiarray failed to import") + */ + __pyx_t_4 = _import_array(); if (unlikely(__pyx_t_4 == ((int)-1))) __PYX_ERR(1, 942, __pyx_L3_error) + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":941 + * # Cython code. + * cdef inline int import_array() except -1: + * try: # <<<<<<<<<<<<<< + * __pyx_import_array() + * except Exception: + */ + } + __Pyx_XDECREF(__pyx_t_1); __pyx_t_1 = 0; + __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0; + __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0; + goto __pyx_L8_try_end; + __pyx_L3_error:; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":943 + * try: + * __pyx_import_array() + * except Exception: # <<<<<<<<<<<<<< + * raise ImportError("numpy.core.multiarray failed to import") + * + */ + __pyx_t_4 = __Pyx_PyErr_ExceptionMatches(((PyObject *)(&((PyTypeObject*)PyExc_Exception)[0]))); + if (__pyx_t_4) { + __Pyx_AddTraceback("numpy.import_array", __pyx_clineno, __pyx_lineno, __pyx_filename); + if (__Pyx_GetException(&__pyx_t_5, &__pyx_t_6, &__pyx_t_7) < 0) __PYX_ERR(1, 943, __pyx_L5_except_error) + __Pyx_GOTREF(__pyx_t_5); + __Pyx_GOTREF(__pyx_t_6); + __Pyx_GOTREF(__pyx_t_7); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":944 + * __pyx_import_array() + * except Exception: + * raise ImportError("numpy.core.multiarray failed to import") # <<<<<<<<<<<<<< + * + * cdef inline int import_umath() except -1: + */ + __pyx_t_8 = __Pyx_PyObject_Call(__pyx_builtin_ImportError, __pyx_tuple_, NULL); if (unlikely(!__pyx_t_8)) __PYX_ERR(1, 944, __pyx_L5_except_error) + __Pyx_GOTREF(__pyx_t_8); + __Pyx_Raise(__pyx_t_8, 0, 0, 0); + __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0; + __PYX_ERR(1, 944, __pyx_L5_except_error) + } + goto __pyx_L5_except_error; + __pyx_L5_except_error:; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":941 + * # Cython code. + * cdef inline int import_array() except -1: + * try: # <<<<<<<<<<<<<< + * __pyx_import_array() + * except Exception: + */ + __Pyx_XGIVEREF(__pyx_t_1); + __Pyx_XGIVEREF(__pyx_t_2); + __Pyx_XGIVEREF(__pyx_t_3); + __Pyx_ExceptionReset(__pyx_t_1, __pyx_t_2, __pyx_t_3); + goto __pyx_L1_error; + __pyx_L8_try_end:; + } + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":940 + * # Versions of the import_* functions which are more suitable for + * # Cython code. + * cdef inline int import_array() except -1: # <<<<<<<<<<<<<< + * try: + * __pyx_import_array() + */ + + /* function exit code */ + __pyx_r = 0; + goto __pyx_L0; + __pyx_L1_error:; + __Pyx_XDECREF(__pyx_t_5); + __Pyx_XDECREF(__pyx_t_6); + __Pyx_XDECREF(__pyx_t_7); + __Pyx_XDECREF(__pyx_t_8); + __Pyx_AddTraceback("numpy.import_array", __pyx_clineno, __pyx_lineno, __pyx_filename); + __pyx_r = -1; + __pyx_L0:; + __Pyx_RefNannyFinishContext(); + return __pyx_r; +} + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":946 + * raise ImportError("numpy.core.multiarray failed to import") + * + * cdef inline int import_umath() except -1: # <<<<<<<<<<<<<< + * try: + * _import_umath() + */ + +static CYTHON_INLINE int __pyx_f_5numpy_import_umath(void) { + int __pyx_r; + __Pyx_RefNannyDeclarations + PyObject *__pyx_t_1 = NULL; + PyObject *__pyx_t_2 = NULL; + PyObject *__pyx_t_3 = NULL; + int __pyx_t_4; + PyObject *__pyx_t_5 = NULL; + PyObject *__pyx_t_6 = NULL; + PyObject *__pyx_t_7 = NULL; + PyObject *__pyx_t_8 = NULL; + int __pyx_lineno = 0; + const char *__pyx_filename = NULL; + int __pyx_clineno = 0; + __Pyx_RefNannySetupContext("import_umath", 0); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":947 + * + * cdef inline int import_umath() except -1: + * try: # <<<<<<<<<<<<<< + * _import_umath() + * except Exception: + */ + { + __Pyx_PyThreadState_declare + __Pyx_PyThreadState_assign + __Pyx_ExceptionSave(&__pyx_t_1, &__pyx_t_2, &__pyx_t_3); + __Pyx_XGOTREF(__pyx_t_1); + __Pyx_XGOTREF(__pyx_t_2); + __Pyx_XGOTREF(__pyx_t_3); + /*try:*/ { + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":948 + * cdef inline int import_umath() except -1: + * try: + * _import_umath() # <<<<<<<<<<<<<< + * except Exception: + * raise ImportError("numpy.core.umath failed to import") + */ + __pyx_t_4 = _import_umath(); if (unlikely(__pyx_t_4 == ((int)-1))) __PYX_ERR(1, 948, __pyx_L3_error) + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":947 + * + * cdef inline int import_umath() except -1: + * try: # <<<<<<<<<<<<<< + * _import_umath() + * except Exception: + */ + } + __Pyx_XDECREF(__pyx_t_1); __pyx_t_1 = 0; + __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0; + __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0; + goto __pyx_L8_try_end; + __pyx_L3_error:; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":949 + * try: + * _import_umath() + * except Exception: # <<<<<<<<<<<<<< + * raise ImportError("numpy.core.umath failed to import") + * + */ + __pyx_t_4 = __Pyx_PyErr_ExceptionMatches(((PyObject *)(&((PyTypeObject*)PyExc_Exception)[0]))); + if (__pyx_t_4) { + __Pyx_AddTraceback("numpy.import_umath", __pyx_clineno, __pyx_lineno, __pyx_filename); + if (__Pyx_GetException(&__pyx_t_5, &__pyx_t_6, &__pyx_t_7) < 0) __PYX_ERR(1, 949, __pyx_L5_except_error) + __Pyx_GOTREF(__pyx_t_5); + __Pyx_GOTREF(__pyx_t_6); + __Pyx_GOTREF(__pyx_t_7); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":950 + * _import_umath() + * except Exception: + * raise ImportError("numpy.core.umath failed to import") # <<<<<<<<<<<<<< + * + * cdef inline int import_ufunc() except -1: + */ + __pyx_t_8 = __Pyx_PyObject_Call(__pyx_builtin_ImportError, __pyx_tuple__2, NULL); if (unlikely(!__pyx_t_8)) __PYX_ERR(1, 950, __pyx_L5_except_error) + __Pyx_GOTREF(__pyx_t_8); + __Pyx_Raise(__pyx_t_8, 0, 0, 0); + __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0; + __PYX_ERR(1, 950, __pyx_L5_except_error) + } + goto __pyx_L5_except_error; + __pyx_L5_except_error:; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":947 + * + * cdef inline int import_umath() except -1: + * try: # <<<<<<<<<<<<<< + * _import_umath() + * except Exception: + */ + __Pyx_XGIVEREF(__pyx_t_1); + __Pyx_XGIVEREF(__pyx_t_2); + __Pyx_XGIVEREF(__pyx_t_3); + __Pyx_ExceptionReset(__pyx_t_1, __pyx_t_2, __pyx_t_3); + goto __pyx_L1_error; + __pyx_L8_try_end:; + } + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":946 + * raise ImportError("numpy.core.multiarray failed to import") + * + * cdef inline int import_umath() except -1: # <<<<<<<<<<<<<< + * try: + * _import_umath() + */ + + /* function exit code */ + __pyx_r = 0; + goto __pyx_L0; + __pyx_L1_error:; + __Pyx_XDECREF(__pyx_t_5); + __Pyx_XDECREF(__pyx_t_6); + __Pyx_XDECREF(__pyx_t_7); + __Pyx_XDECREF(__pyx_t_8); + __Pyx_AddTraceback("numpy.import_umath", __pyx_clineno, __pyx_lineno, __pyx_filename); + __pyx_r = -1; + __pyx_L0:; + __Pyx_RefNannyFinishContext(); + return __pyx_r; +} + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":952 + * raise ImportError("numpy.core.umath failed to import") + * + * cdef inline int import_ufunc() except -1: # <<<<<<<<<<<<<< + * try: + * _import_umath() + */ + +static CYTHON_INLINE int __pyx_f_5numpy_import_ufunc(void) { + int __pyx_r; + __Pyx_RefNannyDeclarations + PyObject *__pyx_t_1 = NULL; + PyObject *__pyx_t_2 = NULL; + PyObject *__pyx_t_3 = NULL; + int __pyx_t_4; + PyObject *__pyx_t_5 = NULL; + PyObject *__pyx_t_6 = NULL; + PyObject *__pyx_t_7 = NULL; + PyObject *__pyx_t_8 = NULL; + int __pyx_lineno = 0; + const char *__pyx_filename = NULL; + int __pyx_clineno = 0; + __Pyx_RefNannySetupContext("import_ufunc", 0); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":953 + * + * cdef inline int import_ufunc() except -1: + * try: # <<<<<<<<<<<<<< + * _import_umath() + * except Exception: + */ + { + __Pyx_PyThreadState_declare + __Pyx_PyThreadState_assign + __Pyx_ExceptionSave(&__pyx_t_1, &__pyx_t_2, &__pyx_t_3); + __Pyx_XGOTREF(__pyx_t_1); + __Pyx_XGOTREF(__pyx_t_2); + __Pyx_XGOTREF(__pyx_t_3); + /*try:*/ { + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":954 + * cdef inline int import_ufunc() except -1: + * try: + * _import_umath() # <<<<<<<<<<<<<< + * except Exception: + * raise ImportError("numpy.core.umath failed to import") + */ + __pyx_t_4 = _import_umath(); if (unlikely(__pyx_t_4 == ((int)-1))) __PYX_ERR(1, 954, __pyx_L3_error) + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":953 + * + * cdef inline int import_ufunc() except -1: + * try: # <<<<<<<<<<<<<< + * _import_umath() + * except Exception: + */ + } + __Pyx_XDECREF(__pyx_t_1); __pyx_t_1 = 0; + __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0; + __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0; + goto __pyx_L8_try_end; + __pyx_L3_error:; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":955 + * try: + * _import_umath() + * except Exception: # <<<<<<<<<<<<<< + * raise ImportError("numpy.core.umath failed to import") + * + */ + __pyx_t_4 = __Pyx_PyErr_ExceptionMatches(((PyObject *)(&((PyTypeObject*)PyExc_Exception)[0]))); + if (__pyx_t_4) { + __Pyx_AddTraceback("numpy.import_ufunc", __pyx_clineno, __pyx_lineno, __pyx_filename); + if (__Pyx_GetException(&__pyx_t_5, &__pyx_t_6, &__pyx_t_7) < 0) __PYX_ERR(1, 955, __pyx_L5_except_error) + __Pyx_GOTREF(__pyx_t_5); + __Pyx_GOTREF(__pyx_t_6); + __Pyx_GOTREF(__pyx_t_7); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":956 + * _import_umath() + * except Exception: + * raise ImportError("numpy.core.umath failed to import") # <<<<<<<<<<<<<< + * + * cdef extern from *: + */ + __pyx_t_8 = __Pyx_PyObject_Call(__pyx_builtin_ImportError, __pyx_tuple__2, NULL); if (unlikely(!__pyx_t_8)) __PYX_ERR(1, 956, __pyx_L5_except_error) + __Pyx_GOTREF(__pyx_t_8); + __Pyx_Raise(__pyx_t_8, 0, 0, 0); + __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0; + __PYX_ERR(1, 956, __pyx_L5_except_error) + } + goto __pyx_L5_except_error; + __pyx_L5_except_error:; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":953 + * + * cdef inline int import_ufunc() except -1: + * try: # <<<<<<<<<<<<<< + * _import_umath() + * except Exception: + */ + __Pyx_XGIVEREF(__pyx_t_1); + __Pyx_XGIVEREF(__pyx_t_2); + __Pyx_XGIVEREF(__pyx_t_3); + __Pyx_ExceptionReset(__pyx_t_1, __pyx_t_2, __pyx_t_3); + goto __pyx_L1_error; + __pyx_L8_try_end:; + } + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":952 + * raise ImportError("numpy.core.umath failed to import") + * + * cdef inline int import_ufunc() except -1: # <<<<<<<<<<<<<< + * try: + * _import_umath() + */ + + /* function exit code */ + __pyx_r = 0; + goto __pyx_L0; + __pyx_L1_error:; + __Pyx_XDECREF(__pyx_t_5); + __Pyx_XDECREF(__pyx_t_6); + __Pyx_XDECREF(__pyx_t_7); + __Pyx_XDECREF(__pyx_t_8); + __Pyx_AddTraceback("numpy.import_ufunc", __pyx_clineno, __pyx_lineno, __pyx_filename); + __pyx_r = -1; + __pyx_L0:; + __Pyx_RefNannyFinishContext(); + return __pyx_r; +} + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":966 + * + * + * cdef inline bint is_timedelta64_object(object obj): # <<<<<<<<<<<<<< + * """ + * Cython equivalent of `isinstance(obj, np.timedelta64)` + */ + +static CYTHON_INLINE int __pyx_f_5numpy_is_timedelta64_object(PyObject *__pyx_v_obj) { + int __pyx_r; + __Pyx_RefNannyDeclarations + __Pyx_RefNannySetupContext("is_timedelta64_object", 0); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":978 + * bool + * """ + * return PyObject_TypeCheck(obj, &PyTimedeltaArrType_Type) # <<<<<<<<<<<<<< + * + * + */ + __pyx_r = PyObject_TypeCheck(__pyx_v_obj, (&PyTimedeltaArrType_Type)); + goto __pyx_L0; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":966 + * + * + * cdef inline bint is_timedelta64_object(object obj): # <<<<<<<<<<<<<< + * """ + * Cython equivalent of `isinstance(obj, np.timedelta64)` + */ + + /* function exit code */ + __pyx_L0:; + __Pyx_RefNannyFinishContext(); + return __pyx_r; +} + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":981 + * + * + * cdef inline bint is_datetime64_object(object obj): # <<<<<<<<<<<<<< + * """ + * Cython equivalent of `isinstance(obj, np.datetime64)` + */ + +static CYTHON_INLINE int __pyx_f_5numpy_is_datetime64_object(PyObject *__pyx_v_obj) { + int __pyx_r; + __Pyx_RefNannyDeclarations + __Pyx_RefNannySetupContext("is_datetime64_object", 0); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":993 + * bool + * """ + * return PyObject_TypeCheck(obj, &PyDatetimeArrType_Type) # <<<<<<<<<<<<<< + * + * + */ + __pyx_r = PyObject_TypeCheck(__pyx_v_obj, (&PyDatetimeArrType_Type)); + goto __pyx_L0; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":981 + * + * + * cdef inline bint is_datetime64_object(object obj): # <<<<<<<<<<<<<< + * """ + * Cython equivalent of `isinstance(obj, np.datetime64)` + */ + + /* function exit code */ + __pyx_L0:; + __Pyx_RefNannyFinishContext(); + return __pyx_r; +} + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":996 + * + * + * cdef inline npy_datetime get_datetime64_value(object obj) nogil: # <<<<<<<<<<<<<< + * """ + * returns the int64 value underlying scalar numpy datetime64 object + */ + +static CYTHON_INLINE npy_datetime __pyx_f_5numpy_get_datetime64_value(PyObject *__pyx_v_obj) { + npy_datetime __pyx_r; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":1003 + * also needed. That can be found using `get_datetime64_unit`. + * """ + * return (obj).obval # <<<<<<<<<<<<<< + * + * + */ + __pyx_r = ((PyDatetimeScalarObject *)__pyx_v_obj)->obval; + goto __pyx_L0; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":996 + * + * + * cdef inline npy_datetime get_datetime64_value(object obj) nogil: # <<<<<<<<<<<<<< + * """ + * returns the int64 value underlying scalar numpy datetime64 object + */ + + /* function exit code */ + __pyx_L0:; + return __pyx_r; +} + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":1006 + * + * + * cdef inline npy_timedelta get_timedelta64_value(object obj) nogil: # <<<<<<<<<<<<<< + * """ + * returns the int64 value underlying scalar numpy timedelta64 object + */ + +static CYTHON_INLINE npy_timedelta __pyx_f_5numpy_get_timedelta64_value(PyObject *__pyx_v_obj) { + npy_timedelta __pyx_r; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":1010 + * returns the int64 value underlying scalar numpy timedelta64 object + * """ + * return (obj).obval # <<<<<<<<<<<<<< + * + * + */ + __pyx_r = ((PyTimedeltaScalarObject *)__pyx_v_obj)->obval; + goto __pyx_L0; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":1006 + * + * + * cdef inline npy_timedelta get_timedelta64_value(object obj) nogil: # <<<<<<<<<<<<<< + * """ + * returns the int64 value underlying scalar numpy timedelta64 object + */ + + /* function exit code */ + __pyx_L0:; + return __pyx_r; +} + +/* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":1013 + * + * + * cdef inline NPY_DATETIMEUNIT get_datetime64_unit(object obj) nogil: # <<<<<<<<<<<<<< + * """ + * returns the unit part of the dtype for a numpy datetime64 object. + */ + +static CYTHON_INLINE NPY_DATETIMEUNIT __pyx_f_5numpy_get_datetime64_unit(PyObject *__pyx_v_obj) { + NPY_DATETIMEUNIT __pyx_r; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":1017 + * returns the unit part of the dtype for a numpy datetime64 object. + * """ + * return (obj).obmeta.base # <<<<<<<<<<<<<< + */ + __pyx_r = ((NPY_DATETIMEUNIT)((PyDatetimeScalarObject *)__pyx_v_obj)->obmeta.base); + goto __pyx_L0; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":1013 + * + * + * cdef inline NPY_DATETIMEUNIT get_datetime64_unit(object obj) nogil: # <<<<<<<<<<<<<< + * """ + * returns the unit part of the dtype for a numpy datetime64 object. + */ + + /* function exit code */ + __pyx_L0:; + return __pyx_r; +} + +static PyMethodDef __pyx_methods[] = { + {0, 0, 0, 0} +}; + +#if PY_MAJOR_VERSION >= 3 +#if CYTHON_PEP489_MULTI_PHASE_INIT +static PyObject* __pyx_pymod_create(PyObject *spec, PyModuleDef *def); /*proto*/ +static int __pyx_pymod_exec_bbox(PyObject* module); /*proto*/ +static PyModuleDef_Slot __pyx_moduledef_slots[] = { + {Py_mod_create, (void*)__pyx_pymod_create}, + {Py_mod_exec, (void*)__pyx_pymod_exec_bbox}, + {0, NULL} +}; +#endif + +static struct PyModuleDef __pyx_moduledef = { + PyModuleDef_HEAD_INIT, + "bbox", + 0, /* m_doc */ + #if CYTHON_PEP489_MULTI_PHASE_INIT + 0, /* m_size */ + #else + -1, /* m_size */ + #endif + __pyx_methods /* m_methods */, + #if CYTHON_PEP489_MULTI_PHASE_INIT + __pyx_moduledef_slots, /* m_slots */ + #else + NULL, /* m_reload */ + #endif + NULL, /* m_traverse */ + NULL, /* m_clear */ + NULL /* m_free */ +}; +#endif +#ifndef CYTHON_SMALL_CODE +#if defined(__clang__) + #define CYTHON_SMALL_CODE +#elif defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 3)) + #define CYTHON_SMALL_CODE __attribute__((cold)) +#else + #define CYTHON_SMALL_CODE +#endif +#endif + +static __Pyx_StringTabEntry __pyx_string_tab[] = { + {&__pyx_n_s_DTYPE, __pyx_k_DTYPE, sizeof(__pyx_k_DTYPE), 0, 0, 1, 1}, + {&__pyx_n_s_ImportError, __pyx_k_ImportError, sizeof(__pyx_k_ImportError), 0, 0, 1, 1}, + {&__pyx_n_s_K, __pyx_k_K, sizeof(__pyx_k_K), 0, 0, 1, 1}, + {&__pyx_n_s_N, __pyx_k_N, sizeof(__pyx_k_N), 0, 0, 1, 1}, + {&__pyx_n_s_bbox, __pyx_k_bbox, sizeof(__pyx_k_bbox), 0, 0, 1, 1}, + {&__pyx_n_s_bbox_overlaps, __pyx_k_bbox_overlaps, sizeof(__pyx_k_bbox_overlaps), 0, 0, 1, 1}, + {&__pyx_n_s_box_area, __pyx_k_box_area, sizeof(__pyx_k_box_area), 0, 0, 1, 1}, + {&__pyx_kp_s_box_overlaps_pyx, __pyx_k_box_overlaps_pyx, sizeof(__pyx_k_box_overlaps_pyx), 0, 0, 1, 0}, + {&__pyx_n_s_boxes, __pyx_k_boxes, sizeof(__pyx_k_boxes), 0, 0, 1, 1}, + {&__pyx_n_s_cline_in_traceback, __pyx_k_cline_in_traceback, sizeof(__pyx_k_cline_in_traceback), 0, 0, 1, 1}, + {&__pyx_n_s_dtype, __pyx_k_dtype, sizeof(__pyx_k_dtype), 0, 0, 1, 1}, + {&__pyx_n_s_float, __pyx_k_float, sizeof(__pyx_k_float), 0, 0, 1, 1}, + {&__pyx_n_s_ih, __pyx_k_ih, sizeof(__pyx_k_ih), 0, 0, 1, 1}, + {&__pyx_n_s_import, __pyx_k_import, sizeof(__pyx_k_import), 0, 0, 1, 1}, + {&__pyx_n_s_iw, __pyx_k_iw, sizeof(__pyx_k_iw), 0, 0, 1, 1}, + {&__pyx_n_s_k, __pyx_k_k, sizeof(__pyx_k_k), 0, 0, 1, 1}, + {&__pyx_n_s_main, __pyx_k_main, sizeof(__pyx_k_main), 0, 0, 1, 1}, + {&__pyx_n_s_n, __pyx_k_n, sizeof(__pyx_k_n), 0, 0, 1, 1}, + {&__pyx_n_s_name, __pyx_k_name, sizeof(__pyx_k_name), 0, 0, 1, 1}, + {&__pyx_n_s_np, __pyx_k_np, sizeof(__pyx_k_np), 0, 0, 1, 1}, + {&__pyx_n_s_numpy, __pyx_k_numpy, sizeof(__pyx_k_numpy), 0, 0, 1, 1}, + {&__pyx_kp_s_numpy_core_multiarray_failed_to, __pyx_k_numpy_core_multiarray_failed_to, sizeof(__pyx_k_numpy_core_multiarray_failed_to), 0, 0, 1, 0}, + {&__pyx_kp_s_numpy_core_umath_failed_to_impor, __pyx_k_numpy_core_umath_failed_to_impor, sizeof(__pyx_k_numpy_core_umath_failed_to_impor), 0, 0, 1, 0}, + {&__pyx_n_s_overlaps, __pyx_k_overlaps, sizeof(__pyx_k_overlaps), 0, 0, 1, 1}, + {&__pyx_n_s_query_boxes, __pyx_k_query_boxes, sizeof(__pyx_k_query_boxes), 0, 0, 1, 1}, + {&__pyx_n_s_range, __pyx_k_range, sizeof(__pyx_k_range), 0, 0, 1, 1}, + {&__pyx_n_s_test, __pyx_k_test, sizeof(__pyx_k_test), 0, 0, 1, 1}, + {&__pyx_n_s_ua, __pyx_k_ua, sizeof(__pyx_k_ua), 0, 0, 1, 1}, + {&__pyx_n_s_zeros, __pyx_k_zeros, sizeof(__pyx_k_zeros), 0, 0, 1, 1}, + {0, 0, 0, 0, 0, 0, 0} +}; +static CYTHON_SMALL_CODE int __Pyx_InitCachedBuiltins(void) { + __pyx_builtin_range = __Pyx_GetBuiltinName(__pyx_n_s_range); if (!__pyx_builtin_range) __PYX_ERR(0, 33, __pyx_L1_error) + __pyx_builtin_ImportError = __Pyx_GetBuiltinName(__pyx_n_s_ImportError); if (!__pyx_builtin_ImportError) __PYX_ERR(1, 944, __pyx_L1_error) + return 0; + __pyx_L1_error:; + return -1; +} + +static CYTHON_SMALL_CODE int __Pyx_InitCachedConstants(void) { + __Pyx_RefNannyDeclarations + __Pyx_RefNannySetupContext("__Pyx_InitCachedConstants", 0); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":944 + * __pyx_import_array() + * except Exception: + * raise ImportError("numpy.core.multiarray failed to import") # <<<<<<<<<<<<<< + * + * cdef inline int import_umath() except -1: + */ + __pyx_tuple_ = PyTuple_Pack(1, __pyx_kp_s_numpy_core_multiarray_failed_to); if (unlikely(!__pyx_tuple_)) __PYX_ERR(1, 944, __pyx_L1_error) + __Pyx_GOTREF(__pyx_tuple_); + __Pyx_GIVEREF(__pyx_tuple_); + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":950 + * _import_umath() + * except Exception: + * raise ImportError("numpy.core.umath failed to import") # <<<<<<<<<<<<<< + * + * cdef inline int import_ufunc() except -1: + */ + __pyx_tuple__2 = PyTuple_Pack(1, __pyx_kp_s_numpy_core_umath_failed_to_impor); if (unlikely(!__pyx_tuple__2)) __PYX_ERR(1, 950, __pyx_L1_error) + __Pyx_GOTREF(__pyx_tuple__2); + __Pyx_GIVEREF(__pyx_tuple__2); + + /* "box_overlaps.pyx":15 + * ctypedef np.float_t DTYPE_t + * + * def bbox_overlaps( # <<<<<<<<<<<<<< + * np.ndarray[DTYPE_t, ndim=2] boxes, + * np.ndarray[DTYPE_t, ndim=2] query_boxes): + */ + __pyx_tuple__3 = PyTuple_Pack(11, __pyx_n_s_boxes, __pyx_n_s_query_boxes, __pyx_n_s_N, __pyx_n_s_K, __pyx_n_s_overlaps, __pyx_n_s_iw, __pyx_n_s_ih, __pyx_n_s_box_area, __pyx_n_s_ua, __pyx_n_s_k, __pyx_n_s_n); if (unlikely(!__pyx_tuple__3)) __PYX_ERR(0, 15, __pyx_L1_error) + __Pyx_GOTREF(__pyx_tuple__3); + __Pyx_GIVEREF(__pyx_tuple__3); + __pyx_codeobj__4 = (PyObject*)__Pyx_PyCode_New(2, 0, 11, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__3, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_box_overlaps_pyx, __pyx_n_s_bbox_overlaps, 15, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__4)) __PYX_ERR(0, 15, __pyx_L1_error) + __Pyx_RefNannyFinishContext(); + return 0; + __pyx_L1_error:; + __Pyx_RefNannyFinishContext(); + return -1; +} + +static CYTHON_SMALL_CODE int __Pyx_InitGlobals(void) { + if (__Pyx_InitStrings(__pyx_string_tab) < 0) __PYX_ERR(0, 1, __pyx_L1_error) + return 0; + __pyx_L1_error:; + return -1; +} + +static CYTHON_SMALL_CODE int __Pyx_modinit_global_init_code(void); /*proto*/ +static CYTHON_SMALL_CODE int __Pyx_modinit_variable_export_code(void); /*proto*/ +static CYTHON_SMALL_CODE int __Pyx_modinit_function_export_code(void); /*proto*/ +static CYTHON_SMALL_CODE int __Pyx_modinit_type_init_code(void); /*proto*/ +static CYTHON_SMALL_CODE int __Pyx_modinit_type_import_code(void); /*proto*/ +static CYTHON_SMALL_CODE int __Pyx_modinit_variable_import_code(void); /*proto*/ +static CYTHON_SMALL_CODE int __Pyx_modinit_function_import_code(void); /*proto*/ + +static int __Pyx_modinit_global_init_code(void) { + __Pyx_RefNannyDeclarations + __Pyx_RefNannySetupContext("__Pyx_modinit_global_init_code", 0); + /*--- Global init code ---*/ + __Pyx_RefNannyFinishContext(); + return 0; +} + +static int __Pyx_modinit_variable_export_code(void) { + __Pyx_RefNannyDeclarations + __Pyx_RefNannySetupContext("__Pyx_modinit_variable_export_code", 0); + /*--- Variable export code ---*/ + __Pyx_RefNannyFinishContext(); + return 0; +} + +static int __Pyx_modinit_function_export_code(void) { + __Pyx_RefNannyDeclarations + __Pyx_RefNannySetupContext("__Pyx_modinit_function_export_code", 0); + /*--- Function export code ---*/ + __Pyx_RefNannyFinishContext(); + return 0; +} + +static int __Pyx_modinit_type_init_code(void) { + __Pyx_RefNannyDeclarations + __Pyx_RefNannySetupContext("__Pyx_modinit_type_init_code", 0); + /*--- Type init code ---*/ + __Pyx_RefNannyFinishContext(); + return 0; +} + +static int __Pyx_modinit_type_import_code(void) { + __Pyx_RefNannyDeclarations + PyObject *__pyx_t_1 = NULL; + int __pyx_lineno = 0; + const char *__pyx_filename = NULL; + int __pyx_clineno = 0; + __Pyx_RefNannySetupContext("__Pyx_modinit_type_import_code", 0); + /*--- Type import code ---*/ + __pyx_t_1 = PyImport_ImportModule(__Pyx_BUILTIN_MODULE_NAME); if (unlikely(!__pyx_t_1)) __PYX_ERR(2, 9, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_1); + __pyx_ptype_7cpython_4type_type = __Pyx_ImportType_0_29_36(__pyx_t_1, __Pyx_BUILTIN_MODULE_NAME, "type", + #if defined(PYPY_VERSION_NUM) && PYPY_VERSION_NUM < 0x050B0000 + sizeof(PyTypeObject), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyTypeObject), + #else + sizeof(PyHeapTypeObject), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyHeapTypeObject), + #endif + __Pyx_ImportType_CheckSize_Warn_0_29_36); if (!__pyx_ptype_7cpython_4type_type) __PYX_ERR(2, 9, __pyx_L1_error) + __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0; + __pyx_t_1 = PyImport_ImportModule("numpy"); if (unlikely(!__pyx_t_1)) __PYX_ERR(1, 199, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_1); + __pyx_ptype_5numpy_dtype = __Pyx_ImportType_0_29_36(__pyx_t_1, "numpy", "dtype", sizeof(PyArray_Descr), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyArray_Descr),__Pyx_ImportType_CheckSize_Ignore_0_29_36); if (!__pyx_ptype_5numpy_dtype) __PYX_ERR(1, 199, __pyx_L1_error) + __pyx_ptype_5numpy_flatiter = __Pyx_ImportType_0_29_36(__pyx_t_1, "numpy", "flatiter", sizeof(PyArrayIterObject), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyArrayIterObject),__Pyx_ImportType_CheckSize_Ignore_0_29_36); if (!__pyx_ptype_5numpy_flatiter) __PYX_ERR(1, 222, __pyx_L1_error) + __pyx_ptype_5numpy_broadcast = __Pyx_ImportType_0_29_36(__pyx_t_1, "numpy", "broadcast", sizeof(PyArrayMultiIterObject), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyArrayMultiIterObject),__Pyx_ImportType_CheckSize_Ignore_0_29_36); if (!__pyx_ptype_5numpy_broadcast) __PYX_ERR(1, 226, __pyx_L1_error) + __pyx_ptype_5numpy_ndarray = __Pyx_ImportType_0_29_36(__pyx_t_1, "numpy", "ndarray", sizeof(PyArrayObject), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyArrayObject),__Pyx_ImportType_CheckSize_Ignore_0_29_36); if (!__pyx_ptype_5numpy_ndarray) __PYX_ERR(1, 238, __pyx_L1_error) + __pyx_ptype_5numpy_generic = __Pyx_ImportType_0_29_36(__pyx_t_1, "numpy", "generic", sizeof(PyObject), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyObject),__Pyx_ImportType_CheckSize_Warn_0_29_36); if (!__pyx_ptype_5numpy_generic) __PYX_ERR(1, 770, __pyx_L1_error) + __pyx_ptype_5numpy_number = __Pyx_ImportType_0_29_36(__pyx_t_1, "numpy", "number", sizeof(PyObject), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyObject),__Pyx_ImportType_CheckSize_Warn_0_29_36); if (!__pyx_ptype_5numpy_number) __PYX_ERR(1, 772, __pyx_L1_error) + __pyx_ptype_5numpy_integer = __Pyx_ImportType_0_29_36(__pyx_t_1, "numpy", "integer", sizeof(PyObject), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyObject),__Pyx_ImportType_CheckSize_Warn_0_29_36); if (!__pyx_ptype_5numpy_integer) __PYX_ERR(1, 774, __pyx_L1_error) + __pyx_ptype_5numpy_signedinteger = __Pyx_ImportType_0_29_36(__pyx_t_1, "numpy", "signedinteger", sizeof(PyObject), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyObject),__Pyx_ImportType_CheckSize_Warn_0_29_36); if (!__pyx_ptype_5numpy_signedinteger) __PYX_ERR(1, 776, __pyx_L1_error) + __pyx_ptype_5numpy_unsignedinteger = __Pyx_ImportType_0_29_36(__pyx_t_1, "numpy", "unsignedinteger", sizeof(PyObject), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyObject),__Pyx_ImportType_CheckSize_Warn_0_29_36); if (!__pyx_ptype_5numpy_unsignedinteger) __PYX_ERR(1, 778, __pyx_L1_error) + __pyx_ptype_5numpy_inexact = __Pyx_ImportType_0_29_36(__pyx_t_1, "numpy", "inexact", sizeof(PyObject), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyObject),__Pyx_ImportType_CheckSize_Warn_0_29_36); if (!__pyx_ptype_5numpy_inexact) __PYX_ERR(1, 780, __pyx_L1_error) + __pyx_ptype_5numpy_floating = __Pyx_ImportType_0_29_36(__pyx_t_1, "numpy", "floating", sizeof(PyObject), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyObject),__Pyx_ImportType_CheckSize_Warn_0_29_36); if (!__pyx_ptype_5numpy_floating) __PYX_ERR(1, 782, __pyx_L1_error) + __pyx_ptype_5numpy_complexfloating = __Pyx_ImportType_0_29_36(__pyx_t_1, "numpy", "complexfloating", sizeof(PyObject), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyObject),__Pyx_ImportType_CheckSize_Warn_0_29_36); if (!__pyx_ptype_5numpy_complexfloating) __PYX_ERR(1, 784, __pyx_L1_error) + __pyx_ptype_5numpy_flexible = __Pyx_ImportType_0_29_36(__pyx_t_1, "numpy", "flexible", sizeof(PyObject), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyObject),__Pyx_ImportType_CheckSize_Warn_0_29_36); if (!__pyx_ptype_5numpy_flexible) __PYX_ERR(1, 786, __pyx_L1_error) + __pyx_ptype_5numpy_character = __Pyx_ImportType_0_29_36(__pyx_t_1, "numpy", "character", sizeof(PyObject), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyObject),__Pyx_ImportType_CheckSize_Warn_0_29_36); if (!__pyx_ptype_5numpy_character) __PYX_ERR(1, 788, __pyx_L1_error) + __pyx_ptype_5numpy_ufunc = __Pyx_ImportType_0_29_36(__pyx_t_1, "numpy", "ufunc", sizeof(PyUFuncObject), __PYX_GET_STRUCT_ALIGNMENT_0_29_36(PyUFuncObject),__Pyx_ImportType_CheckSize_Ignore_0_29_36); if (!__pyx_ptype_5numpy_ufunc) __PYX_ERR(1, 826, __pyx_L1_error) + __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0; + __Pyx_RefNannyFinishContext(); + return 0; + __pyx_L1_error:; + __Pyx_XDECREF(__pyx_t_1); + __Pyx_RefNannyFinishContext(); + return -1; +} + +static int __Pyx_modinit_variable_import_code(void) { + __Pyx_RefNannyDeclarations + __Pyx_RefNannySetupContext("__Pyx_modinit_variable_import_code", 0); + /*--- Variable import code ---*/ + __Pyx_RefNannyFinishContext(); + return 0; +} + +static int __Pyx_modinit_function_import_code(void) { + __Pyx_RefNannyDeclarations + __Pyx_RefNannySetupContext("__Pyx_modinit_function_import_code", 0); + /*--- Function import code ---*/ + __Pyx_RefNannyFinishContext(); + return 0; +} + + +#ifndef CYTHON_NO_PYINIT_EXPORT +#define __Pyx_PyMODINIT_FUNC PyMODINIT_FUNC +#elif PY_MAJOR_VERSION < 3 +#ifdef __cplusplus +#define __Pyx_PyMODINIT_FUNC extern "C" void +#else +#define __Pyx_PyMODINIT_FUNC void +#endif +#else +#ifdef __cplusplus +#define __Pyx_PyMODINIT_FUNC extern "C" PyObject * +#else +#define __Pyx_PyMODINIT_FUNC PyObject * +#endif +#endif + + +#if PY_MAJOR_VERSION < 3 +__Pyx_PyMODINIT_FUNC initbbox(void) CYTHON_SMALL_CODE; /*proto*/ +__Pyx_PyMODINIT_FUNC initbbox(void) +#else +__Pyx_PyMODINIT_FUNC PyInit_bbox(void) CYTHON_SMALL_CODE; /*proto*/ +__Pyx_PyMODINIT_FUNC PyInit_bbox(void) +#if CYTHON_PEP489_MULTI_PHASE_INIT +{ + return PyModuleDef_Init(&__pyx_moduledef); +} +static CYTHON_SMALL_CODE int __Pyx_check_single_interpreter(void) { + #if PY_VERSION_HEX >= 0x030700A1 + static PY_INT64_T main_interpreter_id = -1; + PY_INT64_T current_id = PyInterpreterState_GetID(PyThreadState_Get()->interp); + if (main_interpreter_id == -1) { + main_interpreter_id = current_id; + return (unlikely(current_id == -1)) ? -1 : 0; + } else if (unlikely(main_interpreter_id != current_id)) + #else + static PyInterpreterState *main_interpreter = NULL; + PyInterpreterState *current_interpreter = PyThreadState_Get()->interp; + if (!main_interpreter) { + main_interpreter = current_interpreter; + } else if (unlikely(main_interpreter != current_interpreter)) + #endif + { + PyErr_SetString( + PyExc_ImportError, + "Interpreter change detected - this module can only be loaded into one interpreter per process."); + return -1; + } + return 0; +} +static CYTHON_SMALL_CODE int __Pyx_copy_spec_to_module(PyObject *spec, PyObject *moddict, const char* from_name, const char* to_name, int allow_none) { + PyObject *value = PyObject_GetAttrString(spec, from_name); + int result = 0; + if (likely(value)) { + if (allow_none || value != Py_None) { + result = PyDict_SetItemString(moddict, to_name, value); + } + Py_DECREF(value); + } else if (PyErr_ExceptionMatches(PyExc_AttributeError)) { + PyErr_Clear(); + } else { + result = -1; + } + return result; +} +static CYTHON_SMALL_CODE PyObject* __pyx_pymod_create(PyObject *spec, CYTHON_UNUSED PyModuleDef *def) { + PyObject *module = NULL, *moddict, *modname; + if (__Pyx_check_single_interpreter()) + return NULL; + if (__pyx_m) + return __Pyx_NewRef(__pyx_m); + modname = PyObject_GetAttrString(spec, "name"); + if (unlikely(!modname)) goto bad; + module = PyModule_NewObject(modname); + Py_DECREF(modname); + if (unlikely(!module)) goto bad; + moddict = PyModule_GetDict(module); + if (unlikely(!moddict)) goto bad; + if (unlikely(__Pyx_copy_spec_to_module(spec, moddict, "loader", "__loader__", 1) < 0)) goto bad; + if (unlikely(__Pyx_copy_spec_to_module(spec, moddict, "origin", "__file__", 1) < 0)) goto bad; + if (unlikely(__Pyx_copy_spec_to_module(spec, moddict, "parent", "__package__", 1) < 0)) goto bad; + if (unlikely(__Pyx_copy_spec_to_module(spec, moddict, "submodule_search_locations", "__path__", 0) < 0)) goto bad; + return module; +bad: + Py_XDECREF(module); + return NULL; +} + + +static CYTHON_SMALL_CODE int __pyx_pymod_exec_bbox(PyObject *__pyx_pyinit_module) +#endif +#endif +{ + PyObject *__pyx_t_1 = NULL; + PyObject *__pyx_t_2 = NULL; + int __pyx_lineno = 0; + const char *__pyx_filename = NULL; + int __pyx_clineno = 0; + __Pyx_RefNannyDeclarations + #if CYTHON_PEP489_MULTI_PHASE_INIT + if (__pyx_m) { + if (__pyx_m == __pyx_pyinit_module) return 0; + PyErr_SetString(PyExc_RuntimeError, "Module 'bbox' has already been imported. Re-initialisation is not supported."); + return -1; + } + #elif PY_MAJOR_VERSION >= 3 + if (__pyx_m) return __Pyx_NewRef(__pyx_m); + #endif + #if CYTHON_REFNANNY +__Pyx_RefNanny = __Pyx_RefNannyImportAPI("refnanny"); +if (!__Pyx_RefNanny) { + PyErr_Clear(); + __Pyx_RefNanny = __Pyx_RefNannyImportAPI("Cython.Runtime.refnanny"); + if (!__Pyx_RefNanny) + Py_FatalError("failed to import 'refnanny' module"); +} +#endif + __Pyx_RefNannySetupContext("__Pyx_PyMODINIT_FUNC PyInit_bbox(void)", 0); + if (__Pyx_check_binary_version() < 0) __PYX_ERR(0, 1, __pyx_L1_error) + #ifdef __Pxy_PyFrame_Initialize_Offsets + __Pxy_PyFrame_Initialize_Offsets(); + #endif + __pyx_empty_tuple = PyTuple_New(0); if (unlikely(!__pyx_empty_tuple)) __PYX_ERR(0, 1, __pyx_L1_error) + __pyx_empty_bytes = PyBytes_FromStringAndSize("", 0); if (unlikely(!__pyx_empty_bytes)) __PYX_ERR(0, 1, __pyx_L1_error) + __pyx_empty_unicode = PyUnicode_FromStringAndSize("", 0); if (unlikely(!__pyx_empty_unicode)) __PYX_ERR(0, 1, __pyx_L1_error) + #ifdef __Pyx_CyFunction_USED + if (__pyx_CyFunction_init() < 0) __PYX_ERR(0, 1, __pyx_L1_error) + #endif + #ifdef __Pyx_FusedFunction_USED + if (__pyx_FusedFunction_init() < 0) __PYX_ERR(0, 1, __pyx_L1_error) + #endif + #ifdef __Pyx_Coroutine_USED + if (__pyx_Coroutine_init() < 0) __PYX_ERR(0, 1, __pyx_L1_error) + #endif + #ifdef __Pyx_Generator_USED + if (__pyx_Generator_init() < 0) __PYX_ERR(0, 1, __pyx_L1_error) + #endif + #ifdef __Pyx_AsyncGen_USED + if (__pyx_AsyncGen_init() < 0) __PYX_ERR(0, 1, __pyx_L1_error) + #endif + #ifdef __Pyx_StopAsyncIteration_USED + if (__pyx_StopAsyncIteration_init() < 0) __PYX_ERR(0, 1, __pyx_L1_error) + #endif + /*--- Library function declarations ---*/ + /*--- Threads initialization code ---*/ + #if defined(WITH_THREAD) && PY_VERSION_HEX < 0x030700F0 && defined(__PYX_FORCE_INIT_THREADS) && __PYX_FORCE_INIT_THREADS + PyEval_InitThreads(); + #endif + /*--- Module creation code ---*/ + #if CYTHON_PEP489_MULTI_PHASE_INIT + __pyx_m = __pyx_pyinit_module; + Py_INCREF(__pyx_m); + #else + #if PY_MAJOR_VERSION < 3 + __pyx_m = Py_InitModule4("bbox", __pyx_methods, 0, 0, PYTHON_API_VERSION); Py_XINCREF(__pyx_m); + #else + __pyx_m = PyModule_Create(&__pyx_moduledef); + #endif + if (unlikely(!__pyx_m)) __PYX_ERR(0, 1, __pyx_L1_error) + #endif + __pyx_d = PyModule_GetDict(__pyx_m); if (unlikely(!__pyx_d)) __PYX_ERR(0, 1, __pyx_L1_error) + Py_INCREF(__pyx_d); + __pyx_b = PyImport_AddModule(__Pyx_BUILTIN_MODULE_NAME); if (unlikely(!__pyx_b)) __PYX_ERR(0, 1, __pyx_L1_error) + Py_INCREF(__pyx_b); + __pyx_cython_runtime = PyImport_AddModule((char *) "cython_runtime"); if (unlikely(!__pyx_cython_runtime)) __PYX_ERR(0, 1, __pyx_L1_error) + Py_INCREF(__pyx_cython_runtime); + if (PyObject_SetAttrString(__pyx_m, "__builtins__", __pyx_b) < 0) __PYX_ERR(0, 1, __pyx_L1_error) + /*--- Initialize various global constants etc. ---*/ + if (__Pyx_InitGlobals() < 0) __PYX_ERR(0, 1, __pyx_L1_error) + #if PY_MAJOR_VERSION < 3 && (__PYX_DEFAULT_STRING_ENCODING_IS_ASCII || __PYX_DEFAULT_STRING_ENCODING_IS_DEFAULT) + if (__Pyx_init_sys_getdefaultencoding_params() < 0) __PYX_ERR(0, 1, __pyx_L1_error) + #endif + if (__pyx_module_is_main_bbox) { + if (PyObject_SetAttr(__pyx_m, __pyx_n_s_name, __pyx_n_s_main) < 0) __PYX_ERR(0, 1, __pyx_L1_error) + } + #if PY_MAJOR_VERSION >= 3 + { + PyObject *modules = PyImport_GetModuleDict(); if (unlikely(!modules)) __PYX_ERR(0, 1, __pyx_L1_error) + if (!PyDict_GetItemString(modules, "bbox")) { + if (unlikely(PyDict_SetItemString(modules, "bbox", __pyx_m) < 0)) __PYX_ERR(0, 1, __pyx_L1_error) + } + } + #endif + /*--- Builtin init code ---*/ + if (__Pyx_InitCachedBuiltins() < 0) __PYX_ERR(0, 1, __pyx_L1_error) + /*--- Constants init code ---*/ + if (__Pyx_InitCachedConstants() < 0) __PYX_ERR(0, 1, __pyx_L1_error) + /*--- Global type/function init code ---*/ + (void)__Pyx_modinit_global_init_code(); + (void)__Pyx_modinit_variable_export_code(); + (void)__Pyx_modinit_function_export_code(); + (void)__Pyx_modinit_type_init_code(); + if (unlikely(__Pyx_modinit_type_import_code() < 0)) __PYX_ERR(0, 1, __pyx_L1_error) + (void)__Pyx_modinit_variable_import_code(); + (void)__Pyx_modinit_function_import_code(); + /*--- Execution code ---*/ + #if defined(__Pyx_Generator_USED) || defined(__Pyx_Coroutine_USED) + if (__Pyx_patch_abc() < 0) __PYX_ERR(0, 1, __pyx_L1_error) + #endif + + /* "box_overlaps.pyx":9 + * + * cimport cython + * import numpy as np # <<<<<<<<<<<<<< + * cimport numpy as np + * + */ + __pyx_t_1 = __Pyx_Import(__pyx_n_s_numpy, 0, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 9, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_1); + if (PyDict_SetItem(__pyx_d, __pyx_n_s_np, __pyx_t_1) < 0) __PYX_ERR(0, 9, __pyx_L1_error) + __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0; + + /* "box_overlaps.pyx":12 + * cimport numpy as np + * + * DTYPE = np.float # <<<<<<<<<<<<<< + * ctypedef np.float_t DTYPE_t + * + */ + __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_np); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 12, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_1); + __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_float); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 12, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_2); + __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0; + if (PyDict_SetItem(__pyx_d, __pyx_n_s_DTYPE, __pyx_t_2) < 0) __PYX_ERR(0, 12, __pyx_L1_error) + __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0; + + /* "box_overlaps.pyx":15 + * ctypedef np.float_t DTYPE_t + * + * def bbox_overlaps( # <<<<<<<<<<<<<< + * np.ndarray[DTYPE_t, ndim=2] boxes, + * np.ndarray[DTYPE_t, ndim=2] query_boxes): + */ + __pyx_t_2 = PyCFunction_NewEx(&__pyx_mdef_4bbox_1bbox_overlaps, NULL, __pyx_n_s_bbox); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 15, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_2); + if (PyDict_SetItem(__pyx_d, __pyx_n_s_bbox_overlaps, __pyx_t_2) < 0) __PYX_ERR(0, 15, __pyx_L1_error) + __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0; + + /* "box_overlaps.pyx":1 + * # -------------------------------------------------------- # <<<<<<<<<<<<<< + * # Fast R-CNN + * # Copyright (c) 2015 Microsoft + */ + __pyx_t_2 = __Pyx_PyDict_NewPresized(0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1, __pyx_L1_error) + __Pyx_GOTREF(__pyx_t_2); + if (PyDict_SetItem(__pyx_d, __pyx_n_s_test, __pyx_t_2) < 0) __PYX_ERR(0, 1, __pyx_L1_error) + __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0; + + /* "../../../../../../../../usr/local/lib/python3.10/site-packages/numpy/__init__.pxd":1013 + * + * + * cdef inline NPY_DATETIMEUNIT get_datetime64_unit(object obj) nogil: # <<<<<<<<<<<<<< + * """ + * returns the unit part of the dtype for a numpy datetime64 object. + */ + + /*--- Wrapped vars code ---*/ + + goto __pyx_L0; + __pyx_L1_error:; + __Pyx_XDECREF(__pyx_t_1); + __Pyx_XDECREF(__pyx_t_2); + if (__pyx_m) { + if (__pyx_d) { + __Pyx_AddTraceback("init bbox", __pyx_clineno, __pyx_lineno, __pyx_filename); + } + Py_CLEAR(__pyx_m); + } else if (!PyErr_Occurred()) { + PyErr_SetString(PyExc_ImportError, "init bbox"); + } + __pyx_L0:; + __Pyx_RefNannyFinishContext(); + #if CYTHON_PEP489_MULTI_PHASE_INIT + return (__pyx_m != NULL) ? 0 : -1; + #elif PY_MAJOR_VERSION >= 3 + return __pyx_m; + #else + return; + #endif +} + +/* --- Runtime support code --- */ +/* Refnanny */ +#if CYTHON_REFNANNY +static __Pyx_RefNannyAPIStruct *__Pyx_RefNannyImportAPI(const char *modname) { + PyObject *m = NULL, *p = NULL; + void *r = NULL; + m = PyImport_ImportModule(modname); + if (!m) goto end; + p = PyObject_GetAttrString(m, "RefNannyAPI"); + if (!p) goto end; + r = PyLong_AsVoidPtr(p); +end: + Py_XDECREF(p); + Py_XDECREF(m); + return (__Pyx_RefNannyAPIStruct *)r; +} +#endif + +/* PyObjectGetAttrStr */ +#if CYTHON_USE_TYPE_SLOTS +static CYTHON_INLINE PyObject* __Pyx_PyObject_GetAttrStr(PyObject* obj, PyObject* attr_name) { + PyTypeObject* tp = Py_TYPE(obj); + if (likely(tp->tp_getattro)) + return tp->tp_getattro(obj, attr_name); +#if PY_MAJOR_VERSION < 3 + if (likely(tp->tp_getattr)) + return tp->tp_getattr(obj, PyString_AS_STRING(attr_name)); +#endif + return PyObject_GetAttr(obj, attr_name); +} +#endif + +/* GetBuiltinName */ +static PyObject *__Pyx_GetBuiltinName(PyObject *name) { + PyObject* result = __Pyx_PyObject_GetAttrStr(__pyx_b, name); + if (unlikely(!result)) { + PyErr_Format(PyExc_NameError, +#if PY_MAJOR_VERSION >= 3 + "name '%U' is not defined", name); +#else + "name '%.200s' is not defined", PyString_AS_STRING(name)); +#endif + } + return result; +} + +/* RaiseArgTupleInvalid */ +static void __Pyx_RaiseArgtupleInvalid( + const char* func_name, + int exact, + Py_ssize_t num_min, + Py_ssize_t num_max, + Py_ssize_t num_found) +{ + Py_ssize_t num_expected; + const char *more_or_less; + if (num_found < num_min) { + num_expected = num_min; + more_or_less = "at least"; + } else { + num_expected = num_max; + more_or_less = "at most"; + } + if (exact) { + more_or_less = "exactly"; + } + PyErr_Format(PyExc_TypeError, + "%.200s() takes %.8s %" CYTHON_FORMAT_SSIZE_T "d positional argument%.1s (%" CYTHON_FORMAT_SSIZE_T "d given)", + func_name, more_or_less, num_expected, + (num_expected == 1) ? "" : "s", num_found); +} + +/* RaiseDoubleKeywords */ +static void __Pyx_RaiseDoubleKeywordsError( + const char* func_name, + PyObject* kw_name) +{ + PyErr_Format(PyExc_TypeError, + #if PY_MAJOR_VERSION >= 3 + "%s() got multiple values for keyword argument '%U'", func_name, kw_name); + #else + "%s() got multiple values for keyword argument '%s'", func_name, + PyString_AsString(kw_name)); + #endif +} + +/* ParseKeywords */ +static int __Pyx_ParseOptionalKeywords( + PyObject *kwds, + PyObject **argnames[], + PyObject *kwds2, + PyObject *values[], + Py_ssize_t num_pos_args, + const char* function_name) +{ + PyObject *key = 0, *value = 0; + Py_ssize_t pos = 0; + PyObject*** name; + PyObject*** first_kw_arg = argnames + num_pos_args; + while (PyDict_Next(kwds, &pos, &key, &value)) { + name = first_kw_arg; + while (*name && (**name != key)) name++; + if (*name) { + values[name-argnames] = value; + continue; + } + name = first_kw_arg; + #if PY_MAJOR_VERSION < 3 + if (likely(PyString_Check(key))) { + while (*name) { + if ((CYTHON_COMPILING_IN_PYPY || PyString_GET_SIZE(**name) == PyString_GET_SIZE(key)) + && _PyString_Eq(**name, key)) { + values[name-argnames] = value; + break; + } + name++; + } + if (*name) continue; + else { + PyObject*** argname = argnames; + while (argname != first_kw_arg) { + if ((**argname == key) || ( + (CYTHON_COMPILING_IN_PYPY || PyString_GET_SIZE(**argname) == PyString_GET_SIZE(key)) + && _PyString_Eq(**argname, key))) { + goto arg_passed_twice; + } + argname++; + } + } + } else + #endif + if (likely(PyUnicode_Check(key))) { + while (*name) { + int cmp = (**name == key) ? 0 : + #if !CYTHON_COMPILING_IN_PYPY && PY_MAJOR_VERSION >= 3 + (__Pyx_PyUnicode_GET_LENGTH(**name) != __Pyx_PyUnicode_GET_LENGTH(key)) ? 1 : + #endif + PyUnicode_Compare(**name, key); + if (cmp < 0 && unlikely(PyErr_Occurred())) goto bad; + if (cmp == 0) { + values[name-argnames] = value; + break; + } + name++; + } + if (*name) continue; + else { + PyObject*** argname = argnames; + while (argname != first_kw_arg) { + int cmp = (**argname == key) ? 0 : + #if !CYTHON_COMPILING_IN_PYPY && PY_MAJOR_VERSION >= 3 + (__Pyx_PyUnicode_GET_LENGTH(**argname) != __Pyx_PyUnicode_GET_LENGTH(key)) ? 1 : + #endif + PyUnicode_Compare(**argname, key); + if (cmp < 0 && unlikely(PyErr_Occurred())) goto bad; + if (cmp == 0) goto arg_passed_twice; + argname++; + } + } + } else + goto invalid_keyword_type; + if (kwds2) { + if (unlikely(PyDict_SetItem(kwds2, key, value))) goto bad; + } else { + goto invalid_keyword; + } + } + return 0; +arg_passed_twice: + __Pyx_RaiseDoubleKeywordsError(function_name, key); + goto bad; +invalid_keyword_type: + PyErr_Format(PyExc_TypeError, + "%.200s() keywords must be strings", function_name); + goto bad; +invalid_keyword: + PyErr_Format(PyExc_TypeError, + #if PY_MAJOR_VERSION < 3 + "%.200s() got an unexpected keyword argument '%.200s'", + function_name, PyString_AsString(key)); + #else + "%s() got an unexpected keyword argument '%U'", + function_name, key); + #endif +bad: + return -1; +} + +/* ArgTypeTest */ +static int __Pyx__ArgTypeTest(PyObject *obj, PyTypeObject *type, const char *name, int exact) +{ + if (unlikely(!type)) { + PyErr_SetString(PyExc_SystemError, "Missing type object"); + return 0; + } + else if (exact) { + #if PY_MAJOR_VERSION == 2 + if ((type == &PyBaseString_Type) && likely(__Pyx_PyBaseString_CheckExact(obj))) return 1; + #endif + } + else { + if (likely(__Pyx_TypeCheck(obj, type))) return 1; + } + PyErr_Format(PyExc_TypeError, + "Argument '%.200s' has incorrect type (expected %.200s, got %.200s)", + name, type->tp_name, Py_TYPE(obj)->tp_name); + return 0; +} + +/* IsLittleEndian */ +static CYTHON_INLINE int __Pyx_Is_Little_Endian(void) +{ + union { + uint32_t u32; + uint8_t u8[4]; + } S; + S.u32 = 0x01020304; + return S.u8[0] == 4; +} + +/* BufferFormatCheck */ +static void __Pyx_BufFmt_Init(__Pyx_BufFmt_Context* ctx, + __Pyx_BufFmt_StackElem* stack, + __Pyx_TypeInfo* type) { + stack[0].field = &ctx->root; + stack[0].parent_offset = 0; + ctx->root.type = type; + ctx->root.name = "buffer dtype"; + ctx->root.offset = 0; + ctx->head = stack; + ctx->head->field = &ctx->root; + ctx->fmt_offset = 0; + ctx->head->parent_offset = 0; + ctx->new_packmode = '@'; + ctx->enc_packmode = '@'; + ctx->new_count = 1; + ctx->enc_count = 0; + ctx->enc_type = 0; + ctx->is_complex = 0; + ctx->is_valid_array = 0; + ctx->struct_alignment = 0; + while (type->typegroup == 'S') { + ++ctx->head; + ctx->head->field = type->fields; + ctx->head->parent_offset = 0; + type = type->fields->type; + } +} +static int __Pyx_BufFmt_ParseNumber(const char** ts) { + int count; + const char* t = *ts; + if (*t < '0' || *t > '9') { + return -1; + } else { + count = *t++ - '0'; + while (*t >= '0' && *t <= '9') { + count *= 10; + count += *t++ - '0'; + } + } + *ts = t; + return count; +} +static int __Pyx_BufFmt_ExpectNumber(const char **ts) { + int number = __Pyx_BufFmt_ParseNumber(ts); + if (number == -1) + PyErr_Format(PyExc_ValueError,\ + "Does not understand character buffer dtype format string ('%c')", **ts); + return number; +} +static void __Pyx_BufFmt_RaiseUnexpectedChar(char ch) { + PyErr_Format(PyExc_ValueError, + "Unexpected format string character: '%c'", ch); +} +static const char* __Pyx_BufFmt_DescribeTypeChar(char ch, int is_complex) { + switch (ch) { + case '?': return "'bool'"; + case 'c': return "'char'"; + case 'b': return "'signed char'"; + case 'B': return "'unsigned char'"; + case 'h': return "'short'"; + case 'H': return "'unsigned short'"; + case 'i': return "'int'"; + case 'I': return "'unsigned int'"; + case 'l': return "'long'"; + case 'L': return "'unsigned long'"; + case 'q': return "'long long'"; + case 'Q': return "'unsigned long long'"; + case 'f': return (is_complex ? "'complex float'" : "'float'"); + case 'd': return (is_complex ? "'complex double'" : "'double'"); + case 'g': return (is_complex ? "'complex long double'" : "'long double'"); + case 'T': return "a struct"; + case 'O': return "Python object"; + case 'P': return "a pointer"; + case 's': case 'p': return "a string"; + case 0: return "end"; + default: return "unparseable format string"; + } +} +static size_t __Pyx_BufFmt_TypeCharToStandardSize(char ch, int is_complex) { + switch (ch) { + case '?': case 'c': case 'b': case 'B': case 's': case 'p': return 1; + case 'h': case 'H': return 2; + case 'i': case 'I': case 'l': case 'L': return 4; + case 'q': case 'Q': return 8; + case 'f': return (is_complex ? 8 : 4); + case 'd': return (is_complex ? 16 : 8); + case 'g': { + PyErr_SetString(PyExc_ValueError, "Python does not define a standard format string size for long double ('g').."); + return 0; + } + case 'O': case 'P': return sizeof(void*); + default: + __Pyx_BufFmt_RaiseUnexpectedChar(ch); + return 0; + } +} +static size_t __Pyx_BufFmt_TypeCharToNativeSize(char ch, int is_complex) { + switch (ch) { + case '?': case 'c': case 'b': case 'B': case 's': case 'p': return 1; + case 'h': case 'H': return sizeof(short); + case 'i': case 'I': return sizeof(int); + case 'l': case 'L': return sizeof(long); + #ifdef HAVE_LONG_LONG + case 'q': case 'Q': return sizeof(PY_LONG_LONG); + #endif + case 'f': return sizeof(float) * (is_complex ? 2 : 1); + case 'd': return sizeof(double) * (is_complex ? 2 : 1); + case 'g': return sizeof(long double) * (is_complex ? 2 : 1); + case 'O': case 'P': return sizeof(void*); + default: { + __Pyx_BufFmt_RaiseUnexpectedChar(ch); + return 0; + } + } +} +typedef struct { char c; short x; } __Pyx_st_short; +typedef struct { char c; int x; } __Pyx_st_int; +typedef struct { char c; long x; } __Pyx_st_long; +typedef struct { char c; float x; } __Pyx_st_float; +typedef struct { char c; double x; } __Pyx_st_double; +typedef struct { char c; long double x; } __Pyx_st_longdouble; +typedef struct { char c; void *x; } __Pyx_st_void_p; +#ifdef HAVE_LONG_LONG +typedef struct { char c; PY_LONG_LONG x; } __Pyx_st_longlong; +#endif +static size_t __Pyx_BufFmt_TypeCharToAlignment(char ch, CYTHON_UNUSED int is_complex) { + switch (ch) { + case '?': case 'c': case 'b': case 'B': case 's': case 'p': return 1; + case 'h': case 'H': return sizeof(__Pyx_st_short) - sizeof(short); + case 'i': case 'I': return sizeof(__Pyx_st_int) - sizeof(int); + case 'l': case 'L': return sizeof(__Pyx_st_long) - sizeof(long); +#ifdef HAVE_LONG_LONG + case 'q': case 'Q': return sizeof(__Pyx_st_longlong) - sizeof(PY_LONG_LONG); +#endif + case 'f': return sizeof(__Pyx_st_float) - sizeof(float); + case 'd': return sizeof(__Pyx_st_double) - sizeof(double); + case 'g': return sizeof(__Pyx_st_longdouble) - sizeof(long double); + case 'P': case 'O': return sizeof(__Pyx_st_void_p) - sizeof(void*); + default: + __Pyx_BufFmt_RaiseUnexpectedChar(ch); + return 0; + } +} +/* These are for computing the padding at the end of the struct to align + on the first member of the struct. This will probably the same as above, + but we don't have any guarantees. + */ +typedef struct { short x; char c; } __Pyx_pad_short; +typedef struct { int x; char c; } __Pyx_pad_int; +typedef struct { long x; char c; } __Pyx_pad_long; +typedef struct { float x; char c; } __Pyx_pad_float; +typedef struct { double x; char c; } __Pyx_pad_double; +typedef struct { long double x; char c; } __Pyx_pad_longdouble; +typedef struct { void *x; char c; } __Pyx_pad_void_p; +#ifdef HAVE_LONG_LONG +typedef struct { PY_LONG_LONG x; char c; } __Pyx_pad_longlong; +#endif +static size_t __Pyx_BufFmt_TypeCharToPadding(char ch, CYTHON_UNUSED int is_complex) { + switch (ch) { + case '?': case 'c': case 'b': case 'B': case 's': case 'p': return 1; + case 'h': case 'H': return sizeof(__Pyx_pad_short) - sizeof(short); + case 'i': case 'I': return sizeof(__Pyx_pad_int) - sizeof(int); + case 'l': case 'L': return sizeof(__Pyx_pad_long) - sizeof(long); +#ifdef HAVE_LONG_LONG + case 'q': case 'Q': return sizeof(__Pyx_pad_longlong) - sizeof(PY_LONG_LONG); +#endif + case 'f': return sizeof(__Pyx_pad_float) - sizeof(float); + case 'd': return sizeof(__Pyx_pad_double) - sizeof(double); + case 'g': return sizeof(__Pyx_pad_longdouble) - sizeof(long double); + case 'P': case 'O': return sizeof(__Pyx_pad_void_p) - sizeof(void*); + default: + __Pyx_BufFmt_RaiseUnexpectedChar(ch); + return 0; + } +} +static char __Pyx_BufFmt_TypeCharToGroup(char ch, int is_complex) { + switch (ch) { + case 'c': + return 'H'; + case 'b': case 'h': case 'i': + case 'l': case 'q': case 's': case 'p': + return 'I'; + case '?': case 'B': case 'H': case 'I': case 'L': case 'Q': + return 'U'; + case 'f': case 'd': case 'g': + return (is_complex ? 'C' : 'R'); + case 'O': + return 'O'; + case 'P': + return 'P'; + default: { + __Pyx_BufFmt_RaiseUnexpectedChar(ch); + return 0; + } + } +} +static void __Pyx_BufFmt_RaiseExpected(__Pyx_BufFmt_Context* ctx) { + if (ctx->head == NULL || ctx->head->field == &ctx->root) { + const char* expected; + const char* quote; + if (ctx->head == NULL) { + expected = "end"; + quote = ""; + } else { + expected = ctx->head->field->type->name; + quote = "'"; + } + PyErr_Format(PyExc_ValueError, + "Buffer dtype mismatch, expected %s%s%s but got %s", + quote, expected, quote, + __Pyx_BufFmt_DescribeTypeChar(ctx->enc_type, ctx->is_complex)); + } else { + __Pyx_StructField* field = ctx->head->field; + __Pyx_StructField* parent = (ctx->head - 1)->field; + PyErr_Format(PyExc_ValueError, + "Buffer dtype mismatch, expected '%s' but got %s in '%s.%s'", + field->type->name, __Pyx_BufFmt_DescribeTypeChar(ctx->enc_type, ctx->is_complex), + parent->type->name, field->name); + } +} +static int __Pyx_BufFmt_ProcessTypeChunk(__Pyx_BufFmt_Context* ctx) { + char group; + size_t size, offset, arraysize = 1; + if (ctx->enc_type == 0) return 0; + if (ctx->head->field->type->arraysize[0]) { + int i, ndim = 0; + if (ctx->enc_type == 's' || ctx->enc_type == 'p') { + ctx->is_valid_array = ctx->head->field->type->ndim == 1; + ndim = 1; + if (ctx->enc_count != ctx->head->field->type->arraysize[0]) { + PyErr_Format(PyExc_ValueError, + "Expected a dimension of size %zu, got %zu", + ctx->head->field->type->arraysize[0], ctx->enc_count); + return -1; + } + } + if (!ctx->is_valid_array) { + PyErr_Format(PyExc_ValueError, "Expected %d dimensions, got %d", + ctx->head->field->type->ndim, ndim); + return -1; + } + for (i = 0; i < ctx->head->field->type->ndim; i++) { + arraysize *= ctx->head->field->type->arraysize[i]; + } + ctx->is_valid_array = 0; + ctx->enc_count = 1; + } + group = __Pyx_BufFmt_TypeCharToGroup(ctx->enc_type, ctx->is_complex); + do { + __Pyx_StructField* field = ctx->head->field; + __Pyx_TypeInfo* type = field->type; + if (ctx->enc_packmode == '@' || ctx->enc_packmode == '^') { + size = __Pyx_BufFmt_TypeCharToNativeSize(ctx->enc_type, ctx->is_complex); + } else { + size = __Pyx_BufFmt_TypeCharToStandardSize(ctx->enc_type, ctx->is_complex); + } + if (ctx->enc_packmode == '@') { + size_t align_at = __Pyx_BufFmt_TypeCharToAlignment(ctx->enc_type, ctx->is_complex); + size_t align_mod_offset; + if (align_at == 0) return -1; + align_mod_offset = ctx->fmt_offset % align_at; + if (align_mod_offset > 0) ctx->fmt_offset += align_at - align_mod_offset; + if (ctx->struct_alignment == 0) + ctx->struct_alignment = __Pyx_BufFmt_TypeCharToPadding(ctx->enc_type, + ctx->is_complex); + } + if (type->size != size || type->typegroup != group) { + if (type->typegroup == 'C' && type->fields != NULL) { + size_t parent_offset = ctx->head->parent_offset + field->offset; + ++ctx->head; + ctx->head->field = type->fields; + ctx->head->parent_offset = parent_offset; + continue; + } + if ((type->typegroup == 'H' || group == 'H') && type->size == size) { + } else { + __Pyx_BufFmt_RaiseExpected(ctx); + return -1; + } + } + offset = ctx->head->parent_offset + field->offset; + if (ctx->fmt_offset != offset) { + PyErr_Format(PyExc_ValueError, + "Buffer dtype mismatch; next field is at offset %" CYTHON_FORMAT_SSIZE_T "d but %" CYTHON_FORMAT_SSIZE_T "d expected", + (Py_ssize_t)ctx->fmt_offset, (Py_ssize_t)offset); + return -1; + } + ctx->fmt_offset += size; + if (arraysize) + ctx->fmt_offset += (arraysize - 1) * size; + --ctx->enc_count; + while (1) { + if (field == &ctx->root) { + ctx->head = NULL; + if (ctx->enc_count != 0) { + __Pyx_BufFmt_RaiseExpected(ctx); + return -1; + } + break; + } + ctx->head->field = ++field; + if (field->type == NULL) { + --ctx->head; + field = ctx->head->field; + continue; + } else if (field->type->typegroup == 'S') { + size_t parent_offset = ctx->head->parent_offset + field->offset; + if (field->type->fields->type == NULL) continue; + field = field->type->fields; + ++ctx->head; + ctx->head->field = field; + ctx->head->parent_offset = parent_offset; + break; + } else { + break; + } + } + } while (ctx->enc_count); + ctx->enc_type = 0; + ctx->is_complex = 0; + return 0; +} +static PyObject * +__pyx_buffmt_parse_array(__Pyx_BufFmt_Context* ctx, const char** tsp) +{ + const char *ts = *tsp; + int i = 0, number, ndim; + ++ts; + if (ctx->new_count != 1) { + PyErr_SetString(PyExc_ValueError, + "Cannot handle repeated arrays in format string"); + return NULL; + } + if (__Pyx_BufFmt_ProcessTypeChunk(ctx) == -1) return NULL; + ndim = ctx->head->field->type->ndim; + while (*ts && *ts != ')') { + switch (*ts) { + case ' ': case '\f': case '\r': case '\n': case '\t': case '\v': continue; + default: break; + } + number = __Pyx_BufFmt_ExpectNumber(&ts); + if (number == -1) return NULL; + if (i < ndim && (size_t) number != ctx->head->field->type->arraysize[i]) + return PyErr_Format(PyExc_ValueError, + "Expected a dimension of size %zu, got %d", + ctx->head->field->type->arraysize[i], number); + if (*ts != ',' && *ts != ')') + return PyErr_Format(PyExc_ValueError, + "Expected a comma in format string, got '%c'", *ts); + if (*ts == ',') ts++; + i++; + } + if (i != ndim) + return PyErr_Format(PyExc_ValueError, "Expected %d dimension(s), got %d", + ctx->head->field->type->ndim, i); + if (!*ts) { + PyErr_SetString(PyExc_ValueError, + "Unexpected end of format string, expected ')'"); + return NULL; + } + ctx->is_valid_array = 1; + ctx->new_count = 1; + *tsp = ++ts; + return Py_None; +} +static const char* __Pyx_BufFmt_CheckString(__Pyx_BufFmt_Context* ctx, const char* ts) { + int got_Z = 0; + while (1) { + switch(*ts) { + case 0: + if (ctx->enc_type != 0 && ctx->head == NULL) { + __Pyx_BufFmt_RaiseExpected(ctx); + return NULL; + } + if (__Pyx_BufFmt_ProcessTypeChunk(ctx) == -1) return NULL; + if (ctx->head != NULL) { + __Pyx_BufFmt_RaiseExpected(ctx); + return NULL; + } + return ts; + case ' ': + case '\r': + case '\n': + ++ts; + break; + case '<': + if (!__Pyx_Is_Little_Endian()) { + PyErr_SetString(PyExc_ValueError, "Little-endian buffer not supported on big-endian compiler"); + return NULL; + } + ctx->new_packmode = '='; + ++ts; + break; + case '>': + case '!': + if (__Pyx_Is_Little_Endian()) { + PyErr_SetString(PyExc_ValueError, "Big-endian buffer not supported on little-endian compiler"); + return NULL; + } + ctx->new_packmode = '='; + ++ts; + break; + case '=': + case '@': + case '^': + ctx->new_packmode = *ts++; + break; + case 'T': + { + const char* ts_after_sub; + size_t i, struct_count = ctx->new_count; + size_t struct_alignment = ctx->struct_alignment; + ctx->new_count = 1; + ++ts; + if (*ts != '{') { + PyErr_SetString(PyExc_ValueError, "Buffer acquisition: Expected '{' after 'T'"); + return NULL; + } + if (__Pyx_BufFmt_ProcessTypeChunk(ctx) == -1) return NULL; + ctx->enc_type = 0; + ctx->enc_count = 0; + ctx->struct_alignment = 0; + ++ts; + ts_after_sub = ts; + for (i = 0; i != struct_count; ++i) { + ts_after_sub = __Pyx_BufFmt_CheckString(ctx, ts); + if (!ts_after_sub) return NULL; + } + ts = ts_after_sub; + if (struct_alignment) ctx->struct_alignment = struct_alignment; + } + break; + case '}': + { + size_t alignment = ctx->struct_alignment; + ++ts; + if (__Pyx_BufFmt_ProcessTypeChunk(ctx) == -1) return NULL; + ctx->enc_type = 0; + if (alignment && ctx->fmt_offset % alignment) { + ctx->fmt_offset += alignment - (ctx->fmt_offset % alignment); + } + } + return ts; + case 'x': + if (__Pyx_BufFmt_ProcessTypeChunk(ctx) == -1) return NULL; + ctx->fmt_offset += ctx->new_count; + ctx->new_count = 1; + ctx->enc_count = 0; + ctx->enc_type = 0; + ctx->enc_packmode = ctx->new_packmode; + ++ts; + break; + case 'Z': + got_Z = 1; + ++ts; + if (*ts != 'f' && *ts != 'd' && *ts != 'g') { + __Pyx_BufFmt_RaiseUnexpectedChar('Z'); + return NULL; + } + CYTHON_FALLTHROUGH; + case '?': case 'c': case 'b': case 'B': case 'h': case 'H': case 'i': case 'I': + case 'l': case 'L': case 'q': case 'Q': + case 'f': case 'd': case 'g': + case 'O': case 'p': + if ((ctx->enc_type == *ts) && (got_Z == ctx->is_complex) && + (ctx->enc_packmode == ctx->new_packmode) && (!ctx->is_valid_array)) { + ctx->enc_count += ctx->new_count; + ctx->new_count = 1; + got_Z = 0; + ++ts; + break; + } + CYTHON_FALLTHROUGH; + case 's': + if (__Pyx_BufFmt_ProcessTypeChunk(ctx) == -1) return NULL; + ctx->enc_count = ctx->new_count; + ctx->enc_packmode = ctx->new_packmode; + ctx->enc_type = *ts; + ctx->is_complex = got_Z; + ++ts; + ctx->new_count = 1; + got_Z = 0; + break; + case ':': + ++ts; + while(*ts != ':') ++ts; + ++ts; + break; + case '(': + if (!__pyx_buffmt_parse_array(ctx, &ts)) return NULL; + break; + default: + { + int number = __Pyx_BufFmt_ExpectNumber(&ts); + if (number == -1) return NULL; + ctx->new_count = (size_t)number; + } + } + } +} + +/* BufferGetAndValidate */ + static CYTHON_INLINE void __Pyx_SafeReleaseBuffer(Py_buffer* info) { + if (unlikely(info->buf == NULL)) return; + if (info->suboffsets == __Pyx_minusones) info->suboffsets = NULL; + __Pyx_ReleaseBuffer(info); +} +static void __Pyx_ZeroBuffer(Py_buffer* buf) { + buf->buf = NULL; + buf->obj = NULL; + buf->strides = __Pyx_zeros; + buf->shape = __Pyx_zeros; + buf->suboffsets = __Pyx_minusones; +} +static int __Pyx__GetBufferAndValidate( + Py_buffer* buf, PyObject* obj, __Pyx_TypeInfo* dtype, int flags, + int nd, int cast, __Pyx_BufFmt_StackElem* stack) +{ + buf->buf = NULL; + if (unlikely(__Pyx_GetBuffer(obj, buf, flags) == -1)) { + __Pyx_ZeroBuffer(buf); + return -1; + } + if (unlikely(buf->ndim != nd)) { + PyErr_Format(PyExc_ValueError, + "Buffer has wrong number of dimensions (expected %d, got %d)", + nd, buf->ndim); + goto fail; + } + if (!cast) { + __Pyx_BufFmt_Context ctx; + __Pyx_BufFmt_Init(&ctx, stack, dtype); + if (!__Pyx_BufFmt_CheckString(&ctx, buf->format)) goto fail; + } + if (unlikely((size_t)buf->itemsize != dtype->size)) { + PyErr_Format(PyExc_ValueError, + "Item size of buffer (%" CYTHON_FORMAT_SSIZE_T "d byte%s) does not match size of '%s' (%" CYTHON_FORMAT_SSIZE_T "d byte%s)", + buf->itemsize, (buf->itemsize > 1) ? "s" : "", + dtype->name, (Py_ssize_t)dtype->size, (dtype->size > 1) ? "s" : ""); + goto fail; + } + if (buf->suboffsets == NULL) buf->suboffsets = __Pyx_minusones; + return 0; +fail:; + __Pyx_SafeReleaseBuffer(buf); + return -1; +} + +/* PyDictVersioning */ + #if CYTHON_USE_DICT_VERSIONS && CYTHON_USE_TYPE_SLOTS +static CYTHON_INLINE PY_UINT64_T __Pyx_get_tp_dict_version(PyObject *obj) { + PyObject *dict = Py_TYPE(obj)->tp_dict; + return likely(dict) ? __PYX_GET_DICT_VERSION(dict) : 0; +} +static CYTHON_INLINE PY_UINT64_T __Pyx_get_object_dict_version(PyObject *obj) { + PyObject **dictptr = NULL; + Py_ssize_t offset = Py_TYPE(obj)->tp_dictoffset; + if (offset) { +#if CYTHON_COMPILING_IN_CPYTHON + dictptr = (likely(offset > 0)) ? (PyObject **) ((char *)obj + offset) : _PyObject_GetDictPtr(obj); +#else + dictptr = _PyObject_GetDictPtr(obj); +#endif + } + return (dictptr && *dictptr) ? __PYX_GET_DICT_VERSION(*dictptr) : 0; +} +static CYTHON_INLINE int __Pyx_object_dict_version_matches(PyObject* obj, PY_UINT64_T tp_dict_version, PY_UINT64_T obj_dict_version) { + PyObject *dict = Py_TYPE(obj)->tp_dict; + if (unlikely(!dict) || unlikely(tp_dict_version != __PYX_GET_DICT_VERSION(dict))) + return 0; + return obj_dict_version == __Pyx_get_object_dict_version(obj); +} +#endif + +/* GetModuleGlobalName */ + #if CYTHON_USE_DICT_VERSIONS +static PyObject *__Pyx__GetModuleGlobalName(PyObject *name, PY_UINT64_T *dict_version, PyObject **dict_cached_value) +#else +static CYTHON_INLINE PyObject *__Pyx__GetModuleGlobalName(PyObject *name) +#endif +{ + PyObject *result; +#if !CYTHON_AVOID_BORROWED_REFS +#if CYTHON_COMPILING_IN_CPYTHON && PY_VERSION_HEX >= 0x030500A1 + result = _PyDict_GetItem_KnownHash(__pyx_d, name, ((PyASCIIObject *) name)->hash); + __PYX_UPDATE_DICT_CACHE(__pyx_d, result, *dict_cached_value, *dict_version) + if (likely(result)) { + return __Pyx_NewRef(result); + } else if (unlikely(PyErr_Occurred())) { + return NULL; + } +#else + result = PyDict_GetItem(__pyx_d, name); + __PYX_UPDATE_DICT_CACHE(__pyx_d, result, *dict_cached_value, *dict_version) + if (likely(result)) { + return __Pyx_NewRef(result); + } +#endif +#else + result = PyObject_GetItem(__pyx_d, name); + __PYX_UPDATE_DICT_CACHE(__pyx_d, result, *dict_cached_value, *dict_version) + if (likely(result)) { + return __Pyx_NewRef(result); + } + PyErr_Clear(); +#endif + return __Pyx_GetBuiltinName(name); +} + +/* PyObjectCall */ + #if CYTHON_COMPILING_IN_CPYTHON +static CYTHON_INLINE PyObject* __Pyx_PyObject_Call(PyObject *func, PyObject *arg, PyObject *kw) { + PyObject *result; + ternaryfunc call = Py_TYPE(func)->tp_call; + if (unlikely(!call)) + return PyObject_Call(func, arg, kw); + if (unlikely(Py_EnterRecursiveCall((char*)" while calling a Python object"))) + return NULL; + result = (*call)(func, arg, kw); + Py_LeaveRecursiveCall(); + if (unlikely(!result) && unlikely(!PyErr_Occurred())) { + PyErr_SetString( + PyExc_SystemError, + "NULL result without error in PyObject_Call"); + } + return result; +} +#endif + +/* ExtTypeTest */ + static CYTHON_INLINE int __Pyx_TypeTest(PyObject *obj, PyTypeObject *type) { + if (unlikely(!type)) { + PyErr_SetString(PyExc_SystemError, "Missing type object"); + return 0; + } + if (likely(__Pyx_TypeCheck(obj, type))) + return 1; + PyErr_Format(PyExc_TypeError, "Cannot convert %.200s to %.200s", + Py_TYPE(obj)->tp_name, type->tp_name); + return 0; +} + +/* BufferIndexError */ + static void __Pyx_RaiseBufferIndexError(int axis) { + PyErr_Format(PyExc_IndexError, + "Out of bounds on buffer access (axis %d)", axis); +} + +/* PyErrFetchRestore */ + #if CYTHON_FAST_THREAD_STATE +static CYTHON_INLINE void __Pyx_ErrRestoreInState(PyThreadState *tstate, PyObject *type, PyObject *value, PyObject *tb) { + PyObject *tmp_type, *tmp_value, *tmp_tb; + tmp_type = tstate->curexc_type; + tmp_value = tstate->curexc_value; + tmp_tb = tstate->curexc_traceback; + tstate->curexc_type = type; + tstate->curexc_value = value; + tstate->curexc_traceback = tb; + Py_XDECREF(tmp_type); + Py_XDECREF(tmp_value); + Py_XDECREF(tmp_tb); +} +static CYTHON_INLINE void __Pyx_ErrFetchInState(PyThreadState *tstate, PyObject **type, PyObject **value, PyObject **tb) { + *type = tstate->curexc_type; + *value = tstate->curexc_value; + *tb = tstate->curexc_traceback; + tstate->curexc_type = 0; + tstate->curexc_value = 0; + tstate->curexc_traceback = 0; +} +#endif + +/* GetTopmostException */ + #if CYTHON_USE_EXC_INFO_STACK +static _PyErr_StackItem * +__Pyx_PyErr_GetTopmostException(PyThreadState *tstate) +{ + _PyErr_StackItem *exc_info = tstate->exc_info; + while ((exc_info->exc_type == NULL || exc_info->exc_type == Py_None) && + exc_info->previous_item != NULL) + { + exc_info = exc_info->previous_item; + } + return exc_info; +} +#endif + +/* SaveResetException */ + #if CYTHON_FAST_THREAD_STATE +static CYTHON_INLINE void __Pyx__ExceptionSave(PyThreadState *tstate, PyObject **type, PyObject **value, PyObject **tb) { + #if CYTHON_USE_EXC_INFO_STACK + _PyErr_StackItem *exc_info = __Pyx_PyErr_GetTopmostException(tstate); + *type = exc_info->exc_type; + *value = exc_info->exc_value; + *tb = exc_info->exc_traceback; + #else + *type = tstate->exc_type; + *value = tstate->exc_value; + *tb = tstate->exc_traceback; + #endif + Py_XINCREF(*type); + Py_XINCREF(*value); + Py_XINCREF(*tb); +} +static CYTHON_INLINE void __Pyx__ExceptionReset(PyThreadState *tstate, PyObject *type, PyObject *value, PyObject *tb) { + PyObject *tmp_type, *tmp_value, *tmp_tb; + #if CYTHON_USE_EXC_INFO_STACK + _PyErr_StackItem *exc_info = tstate->exc_info; + tmp_type = exc_info->exc_type; + tmp_value = exc_info->exc_value; + tmp_tb = exc_info->exc_traceback; + exc_info->exc_type = type; + exc_info->exc_value = value; + exc_info->exc_traceback = tb; + #else + tmp_type = tstate->exc_type; + tmp_value = tstate->exc_value; + tmp_tb = tstate->exc_traceback; + tstate->exc_type = type; + tstate->exc_value = value; + tstate->exc_traceback = tb; + #endif + Py_XDECREF(tmp_type); + Py_XDECREF(tmp_value); + Py_XDECREF(tmp_tb); +} +#endif + +/* PyErrExceptionMatches */ + #if CYTHON_FAST_THREAD_STATE +static int __Pyx_PyErr_ExceptionMatchesTuple(PyObject *exc_type, PyObject *tuple) { + Py_ssize_t i, n; + n = PyTuple_GET_SIZE(tuple); +#if PY_MAJOR_VERSION >= 3 + for (i=0; icurexc_type; + if (exc_type == err) return 1; + if (unlikely(!exc_type)) return 0; + if (unlikely(PyTuple_Check(err))) + return __Pyx_PyErr_ExceptionMatchesTuple(exc_type, err); + return __Pyx_PyErr_GivenExceptionMatches(exc_type, err); +} +#endif + +/* GetException */ + #if CYTHON_FAST_THREAD_STATE +static int __Pyx__GetException(PyThreadState *tstate, PyObject **type, PyObject **value, PyObject **tb) +#else +static int __Pyx_GetException(PyObject **type, PyObject **value, PyObject **tb) +#endif +{ + PyObject *local_type, *local_value, *local_tb; +#if CYTHON_FAST_THREAD_STATE + PyObject *tmp_type, *tmp_value, *tmp_tb; + local_type = tstate->curexc_type; + local_value = tstate->curexc_value; + local_tb = tstate->curexc_traceback; + tstate->curexc_type = 0; + tstate->curexc_value = 0; + tstate->curexc_traceback = 0; +#else + PyErr_Fetch(&local_type, &local_value, &local_tb); +#endif + PyErr_NormalizeException(&local_type, &local_value, &local_tb); +#if CYTHON_FAST_THREAD_STATE + if (unlikely(tstate->curexc_type)) +#else + if (unlikely(PyErr_Occurred())) +#endif + goto bad; + #if PY_MAJOR_VERSION >= 3 + if (local_tb) { + if (unlikely(PyException_SetTraceback(local_value, local_tb) < 0)) + goto bad; + } + #endif + Py_XINCREF(local_tb); + Py_XINCREF(local_type); + Py_XINCREF(local_value); + *type = local_type; + *value = local_value; + *tb = local_tb; +#if CYTHON_FAST_THREAD_STATE + #if CYTHON_USE_EXC_INFO_STACK + { + _PyErr_StackItem *exc_info = tstate->exc_info; + tmp_type = exc_info->exc_type; + tmp_value = exc_info->exc_value; + tmp_tb = exc_info->exc_traceback; + exc_info->exc_type = local_type; + exc_info->exc_value = local_value; + exc_info->exc_traceback = local_tb; + } + #else + tmp_type = tstate->exc_type; + tmp_value = tstate->exc_value; + tmp_tb = tstate->exc_traceback; + tstate->exc_type = local_type; + tstate->exc_value = local_value; + tstate->exc_traceback = local_tb; + #endif + Py_XDECREF(tmp_type); + Py_XDECREF(tmp_value); + Py_XDECREF(tmp_tb); +#else + PyErr_SetExcInfo(local_type, local_value, local_tb); +#endif + return 0; +bad: + *type = 0; + *value = 0; + *tb = 0; + Py_XDECREF(local_type); + Py_XDECREF(local_value); + Py_XDECREF(local_tb); + return -1; +} + +/* RaiseException */ + #if PY_MAJOR_VERSION < 3 +static void __Pyx_Raise(PyObject *type, PyObject *value, PyObject *tb, + CYTHON_UNUSED PyObject *cause) { + __Pyx_PyThreadState_declare + Py_XINCREF(type); + if (!value || value == Py_None) + value = NULL; + else + Py_INCREF(value); + if (!tb || tb == Py_None) + tb = NULL; + else { + Py_INCREF(tb); + if (!PyTraceBack_Check(tb)) { + PyErr_SetString(PyExc_TypeError, + "raise: arg 3 must be a traceback or None"); + goto raise_error; + } + } + if (PyType_Check(type)) { +#if CYTHON_COMPILING_IN_PYPY + if (!value) { + Py_INCREF(Py_None); + value = Py_None; + } +#endif + PyErr_NormalizeException(&type, &value, &tb); + } else { + if (value) { + PyErr_SetString(PyExc_TypeError, + "instance exception may not have a separate value"); + goto raise_error; + } + value = type; + type = (PyObject*) Py_TYPE(type); + Py_INCREF(type); + if (!PyType_IsSubtype((PyTypeObject *)type, (PyTypeObject *)PyExc_BaseException)) { + PyErr_SetString(PyExc_TypeError, + "raise: exception class must be a subclass of BaseException"); + goto raise_error; + } + } + __Pyx_PyThreadState_assign + __Pyx_ErrRestore(type, value, tb); + return; +raise_error: + Py_XDECREF(value); + Py_XDECREF(type); + Py_XDECREF(tb); + return; +} +#else +static void __Pyx_Raise(PyObject *type, PyObject *value, PyObject *tb, PyObject *cause) { + PyObject* owned_instance = NULL; + if (tb == Py_None) { + tb = 0; + } else if (tb && !PyTraceBack_Check(tb)) { + PyErr_SetString(PyExc_TypeError, + "raise: arg 3 must be a traceback or None"); + goto bad; + } + if (value == Py_None) + value = 0; + if (PyExceptionInstance_Check(type)) { + if (value) { + PyErr_SetString(PyExc_TypeError, + "instance exception may not have a separate value"); + goto bad; + } + value = type; + type = (PyObject*) Py_TYPE(value); + } else if (PyExceptionClass_Check(type)) { + PyObject *instance_class = NULL; + if (value && PyExceptionInstance_Check(value)) { + instance_class = (PyObject*) Py_TYPE(value); + if (instance_class != type) { + int is_subclass = PyObject_IsSubclass(instance_class, type); + if (!is_subclass) { + instance_class = NULL; + } else if (unlikely(is_subclass == -1)) { + goto bad; + } else { + type = instance_class; + } + } + } + if (!instance_class) { + PyObject *args; + if (!value) + args = PyTuple_New(0); + else if (PyTuple_Check(value)) { + Py_INCREF(value); + args = value; + } else + args = PyTuple_Pack(1, value); + if (!args) + goto bad; + owned_instance = PyObject_Call(type, args, NULL); + Py_DECREF(args); + if (!owned_instance) + goto bad; + value = owned_instance; + if (!PyExceptionInstance_Check(value)) { + PyErr_Format(PyExc_TypeError, + "calling %R should have returned an instance of " + "BaseException, not %R", + type, Py_TYPE(value)); + goto bad; + } + } + } else { + PyErr_SetString(PyExc_TypeError, + "raise: exception class must be a subclass of BaseException"); + goto bad; + } + if (cause) { + PyObject *fixed_cause; + if (cause == Py_None) { + fixed_cause = NULL; + } else if (PyExceptionClass_Check(cause)) { + fixed_cause = PyObject_CallObject(cause, NULL); + if (fixed_cause == NULL) + goto bad; + } else if (PyExceptionInstance_Check(cause)) { + fixed_cause = cause; + Py_INCREF(fixed_cause); + } else { + PyErr_SetString(PyExc_TypeError, + "exception causes must derive from " + "BaseException"); + goto bad; + } + PyException_SetCause(value, fixed_cause); + } + PyErr_SetObject(type, value); + if (tb) { +#if CYTHON_FAST_THREAD_STATE + PyThreadState *tstate = __Pyx_PyThreadState_Current; + PyObject* tmp_tb = tstate->curexc_traceback; + if (tb != tmp_tb) { + Py_INCREF(tb); + tstate->curexc_traceback = tb; + Py_XDECREF(tmp_tb); + } +#else + PyObject *tmp_type, *tmp_value, *tmp_tb; + PyErr_Fetch(&tmp_type, &tmp_value, &tmp_tb); + Py_INCREF(tb); + PyErr_Restore(tmp_type, tmp_value, tb); + Py_XDECREF(tmp_tb); +#endif + } +bad: + Py_XDECREF(owned_instance); + return; +} +#endif + +/* TypeImport */ + #ifndef __PYX_HAVE_RT_ImportType_0_29_36 +#define __PYX_HAVE_RT_ImportType_0_29_36 +static PyTypeObject *__Pyx_ImportType_0_29_36(PyObject *module, const char *module_name, const char *class_name, + size_t size, size_t alignment, enum __Pyx_ImportType_CheckSize_0_29_36 check_size) +{ + PyObject *result = 0; + char warning[200]; + Py_ssize_t basicsize; + Py_ssize_t itemsize; +#ifdef Py_LIMITED_API + PyObject *py_basicsize; + PyObject *py_itemsize; +#endif + result = PyObject_GetAttrString(module, class_name); + if (!result) + goto bad; + if (!PyType_Check(result)) { + PyErr_Format(PyExc_TypeError, + "%.200s.%.200s is not a type object", + module_name, class_name); + goto bad; + } +#ifndef Py_LIMITED_API + basicsize = ((PyTypeObject *)result)->tp_basicsize; + itemsize = ((PyTypeObject *)result)->tp_itemsize; +#else + py_basicsize = PyObject_GetAttrString(result, "__basicsize__"); + if (!py_basicsize) + goto bad; + basicsize = PyLong_AsSsize_t(py_basicsize); + Py_DECREF(py_basicsize); + py_basicsize = 0; + if (basicsize == (Py_ssize_t)-1 && PyErr_Occurred()) + goto bad; + py_itemsize = PyObject_GetAttrString(result, "__itemsize__"); + if (!py_itemsize) + goto bad; + itemsize = PyLong_AsSsize_t(py_itemsize); + Py_DECREF(py_itemsize); + py_itemsize = 0; + if (itemsize == (Py_ssize_t)-1 && PyErr_Occurred()) + goto bad; +#endif + if (itemsize) { + if (size % alignment) { + alignment = size % alignment; + } + if (itemsize < (Py_ssize_t)alignment) + itemsize = (Py_ssize_t)alignment; + } + if ((size_t)(basicsize + itemsize) < size) { + PyErr_Format(PyExc_ValueError, + "%.200s.%.200s size changed, may indicate binary incompatibility. " + "Expected %zd from C header, got %zd from PyObject", + module_name, class_name, size, basicsize); + goto bad; + } + if (check_size == __Pyx_ImportType_CheckSize_Error_0_29_36 && (size_t)basicsize != size) { + PyErr_Format(PyExc_ValueError, + "%.200s.%.200s size changed, may indicate binary incompatibility. " + "Expected %zd from C header, got %zd from PyObject", + module_name, class_name, size, basicsize); + goto bad; + } + else if (check_size == __Pyx_ImportType_CheckSize_Warn_0_29_36 && (size_t)basicsize > size) { + PyOS_snprintf(warning, sizeof(warning), + "%s.%s size changed, may indicate binary incompatibility. " + "Expected %zd from C header, got %zd from PyObject", + module_name, class_name, size, basicsize); + if (PyErr_WarnEx(NULL, warning, 0) < 0) goto bad; + } + return (PyTypeObject *)result; +bad: + Py_XDECREF(result); + return NULL; +} +#endif + +/* Import */ + static PyObject *__Pyx_Import(PyObject *name, PyObject *from_list, int level) { + PyObject *empty_list = 0; + PyObject *module = 0; + PyObject *global_dict = 0; + PyObject *empty_dict = 0; + PyObject *list; + #if PY_MAJOR_VERSION < 3 + PyObject *py_import; + py_import = __Pyx_PyObject_GetAttrStr(__pyx_b, __pyx_n_s_import); + if (!py_import) + goto bad; + #endif + if (from_list) + list = from_list; + else { + empty_list = PyList_New(0); + if (!empty_list) + goto bad; + list = empty_list; + } + global_dict = PyModule_GetDict(__pyx_m); + if (!global_dict) + goto bad; + empty_dict = PyDict_New(); + if (!empty_dict) + goto bad; + { + #if PY_MAJOR_VERSION >= 3 + if (level == -1) { + if ((1) && (strchr(__Pyx_MODULE_NAME, '.'))) { + module = PyImport_ImportModuleLevelObject( + name, global_dict, empty_dict, list, 1); + if (!module) { + if (!PyErr_ExceptionMatches(PyExc_ImportError)) + goto bad; + PyErr_Clear(); + } + } + level = 0; + } + #endif + if (!module) { + #if PY_MAJOR_VERSION < 3 + PyObject *py_level = PyInt_FromLong(level); + if (!py_level) + goto bad; + module = PyObject_CallFunctionObjArgs(py_import, + name, global_dict, empty_dict, list, py_level, (PyObject *)NULL); + Py_DECREF(py_level); + #else + module = PyImport_ImportModuleLevelObject( + name, global_dict, empty_dict, list, level); + #endif + } + } +bad: + #if PY_MAJOR_VERSION < 3 + Py_XDECREF(py_import); + #endif + Py_XDECREF(empty_list); + Py_XDECREF(empty_dict); + return module; +} + +/* CLineInTraceback */ + #ifndef CYTHON_CLINE_IN_TRACEBACK +static int __Pyx_CLineForTraceback(CYTHON_UNUSED PyThreadState *tstate, int c_line) { + PyObject *use_cline; + PyObject *ptype, *pvalue, *ptraceback; +#if CYTHON_COMPILING_IN_CPYTHON + PyObject **cython_runtime_dict; +#endif + if (unlikely(!__pyx_cython_runtime)) { + return c_line; + } + __Pyx_ErrFetchInState(tstate, &ptype, &pvalue, &ptraceback); +#if CYTHON_COMPILING_IN_CPYTHON + cython_runtime_dict = _PyObject_GetDictPtr(__pyx_cython_runtime); + if (likely(cython_runtime_dict)) { + __PYX_PY_DICT_LOOKUP_IF_MODIFIED( + use_cline, *cython_runtime_dict, + __Pyx_PyDict_GetItemStr(*cython_runtime_dict, __pyx_n_s_cline_in_traceback)) + } else +#endif + { + PyObject *use_cline_obj = __Pyx_PyObject_GetAttrStr(__pyx_cython_runtime, __pyx_n_s_cline_in_traceback); + if (use_cline_obj) { + use_cline = PyObject_Not(use_cline_obj) ? Py_False : Py_True; + Py_DECREF(use_cline_obj); + } else { + PyErr_Clear(); + use_cline = NULL; + } + } + if (!use_cline) { + c_line = 0; + (void) PyObject_SetAttr(__pyx_cython_runtime, __pyx_n_s_cline_in_traceback, Py_False); + } + else if (use_cline == Py_False || (use_cline != Py_True && PyObject_Not(use_cline) != 0)) { + c_line = 0; + } + __Pyx_ErrRestoreInState(tstate, ptype, pvalue, ptraceback); + return c_line; +} +#endif + +/* CodeObjectCache */ + static int __pyx_bisect_code_objects(__Pyx_CodeObjectCacheEntry* entries, int count, int code_line) { + int start = 0, mid = 0, end = count - 1; + if (end >= 0 && code_line > entries[end].code_line) { + return count; + } + while (start < end) { + mid = start + (end - start) / 2; + if (code_line < entries[mid].code_line) { + end = mid; + } else if (code_line > entries[mid].code_line) { + start = mid + 1; + } else { + return mid; + } + } + if (code_line <= entries[mid].code_line) { + return mid; + } else { + return mid + 1; + } +} +static PyCodeObject *__pyx_find_code_object(int code_line) { + PyCodeObject* code_object; + int pos; + if (unlikely(!code_line) || unlikely(!__pyx_code_cache.entries)) { + return NULL; + } + pos = __pyx_bisect_code_objects(__pyx_code_cache.entries, __pyx_code_cache.count, code_line); + if (unlikely(pos >= __pyx_code_cache.count) || unlikely(__pyx_code_cache.entries[pos].code_line != code_line)) { + return NULL; + } + code_object = __pyx_code_cache.entries[pos].code_object; + Py_INCREF(code_object); + return code_object; +} +static void __pyx_insert_code_object(int code_line, PyCodeObject* code_object) { + int pos, i; + __Pyx_CodeObjectCacheEntry* entries = __pyx_code_cache.entries; + if (unlikely(!code_line)) { + return; + } + if (unlikely(!entries)) { + entries = (__Pyx_CodeObjectCacheEntry*)PyMem_Malloc(64*sizeof(__Pyx_CodeObjectCacheEntry)); + if (likely(entries)) { + __pyx_code_cache.entries = entries; + __pyx_code_cache.max_count = 64; + __pyx_code_cache.count = 1; + entries[0].code_line = code_line; + entries[0].code_object = code_object; + Py_INCREF(code_object); + } + return; + } + pos = __pyx_bisect_code_objects(__pyx_code_cache.entries, __pyx_code_cache.count, code_line); + if ((pos < __pyx_code_cache.count) && unlikely(__pyx_code_cache.entries[pos].code_line == code_line)) { + PyCodeObject* tmp = entries[pos].code_object; + entries[pos].code_object = code_object; + Py_DECREF(tmp); + return; + } + if (__pyx_code_cache.count == __pyx_code_cache.max_count) { + int new_max = __pyx_code_cache.max_count + 64; + entries = (__Pyx_CodeObjectCacheEntry*)PyMem_Realloc( + __pyx_code_cache.entries, ((size_t)new_max) * sizeof(__Pyx_CodeObjectCacheEntry)); + if (unlikely(!entries)) { + return; + } + __pyx_code_cache.entries = entries; + __pyx_code_cache.max_count = new_max; + } + for (i=__pyx_code_cache.count; i>pos; i--) { + entries[i] = entries[i-1]; + } + entries[pos].code_line = code_line; + entries[pos].code_object = code_object; + __pyx_code_cache.count++; + Py_INCREF(code_object); +} + +/* AddTraceback */ + #include "compile.h" +#include "frameobject.h" +#include "traceback.h" +#if PY_VERSION_HEX >= 0x030b00a6 + #ifndef Py_BUILD_CORE + #define Py_BUILD_CORE 1 + #endif + #include "internal/pycore_frame.h" +#endif +static PyCodeObject* __Pyx_CreateCodeObjectForTraceback( + const char *funcname, int c_line, + int py_line, const char *filename) { + PyCodeObject *py_code = NULL; + PyObject *py_funcname = NULL; + #if PY_MAJOR_VERSION < 3 + PyObject *py_srcfile = NULL; + py_srcfile = PyString_FromString(filename); + if (!py_srcfile) goto bad; + #endif + if (c_line) { + #if PY_MAJOR_VERSION < 3 + py_funcname = PyString_FromFormat( "%s (%s:%d)", funcname, __pyx_cfilenm, c_line); + if (!py_funcname) goto bad; + #else + py_funcname = PyUnicode_FromFormat( "%s (%s:%d)", funcname, __pyx_cfilenm, c_line); + if (!py_funcname) goto bad; + funcname = PyUnicode_AsUTF8(py_funcname); + if (!funcname) goto bad; + #endif + } + else { + #if PY_MAJOR_VERSION < 3 + py_funcname = PyString_FromString(funcname); + if (!py_funcname) goto bad; + #endif + } + #if PY_MAJOR_VERSION < 3 + py_code = __Pyx_PyCode_New( + 0, + 0, + 0, + 0, + 0, + __pyx_empty_bytes, /*PyObject *code,*/ + __pyx_empty_tuple, /*PyObject *consts,*/ + __pyx_empty_tuple, /*PyObject *names,*/ + __pyx_empty_tuple, /*PyObject *varnames,*/ + __pyx_empty_tuple, /*PyObject *freevars,*/ + __pyx_empty_tuple, /*PyObject *cellvars,*/ + py_srcfile, /*PyObject *filename,*/ + py_funcname, /*PyObject *name,*/ + py_line, + __pyx_empty_bytes /*PyObject *lnotab*/ + ); + Py_DECREF(py_srcfile); + #else + py_code = PyCode_NewEmpty(filename, funcname, py_line); + #endif + Py_XDECREF(py_funcname); // XDECREF since it's only set on Py3 if cline + return py_code; +bad: + Py_XDECREF(py_funcname); + #if PY_MAJOR_VERSION < 3 + Py_XDECREF(py_srcfile); + #endif + return NULL; +} +static void __Pyx_AddTraceback(const char *funcname, int c_line, + int py_line, const char *filename) { + PyCodeObject *py_code = 0; + PyFrameObject *py_frame = 0; + PyThreadState *tstate = __Pyx_PyThreadState_Current; + PyObject *ptype, *pvalue, *ptraceback; + if (c_line) { + c_line = __Pyx_CLineForTraceback(tstate, c_line); + } + py_code = __pyx_find_code_object(c_line ? -c_line : py_line); + if (!py_code) { + __Pyx_ErrFetchInState(tstate, &ptype, &pvalue, &ptraceback); + py_code = __Pyx_CreateCodeObjectForTraceback( + funcname, c_line, py_line, filename); + if (!py_code) { + /* If the code object creation fails, then we should clear the + fetched exception references and propagate the new exception */ + Py_XDECREF(ptype); + Py_XDECREF(pvalue); + Py_XDECREF(ptraceback); + goto bad; + } + __Pyx_ErrRestoreInState(tstate, ptype, pvalue, ptraceback); + __pyx_insert_code_object(c_line ? -c_line : py_line, py_code); + } + py_frame = PyFrame_New( + tstate, /*PyThreadState *tstate,*/ + py_code, /*PyCodeObject *code,*/ + __pyx_d, /*PyObject *globals,*/ + 0 /*PyObject *locals*/ + ); + if (!py_frame) goto bad; + __Pyx_PyFrame_SetLineNumber(py_frame, py_line); + PyTraceBack_Here(py_frame); +bad: + Py_XDECREF(py_code); + Py_XDECREF(py_frame); +} + +#if PY_MAJOR_VERSION < 3 +static int __Pyx_GetBuffer(PyObject *obj, Py_buffer *view, int flags) { + if (PyObject_CheckBuffer(obj)) return PyObject_GetBuffer(obj, view, flags); + PyErr_Format(PyExc_TypeError, "'%.200s' does not have the buffer interface", Py_TYPE(obj)->tp_name); + return -1; +} +static void __Pyx_ReleaseBuffer(Py_buffer *view) { + PyObject *obj = view->obj; + if (!obj) return; + if (PyObject_CheckBuffer(obj)) { + PyBuffer_Release(view); + return; + } + if ((0)) {} + view->obj = NULL; + Py_DECREF(obj); +} +#endif + + + /* CIntFromPyVerify */ + #define __PYX_VERIFY_RETURN_INT(target_type, func_type, func_value)\ + __PYX__VERIFY_RETURN_INT(target_type, func_type, func_value, 0) +#define __PYX_VERIFY_RETURN_INT_EXC(target_type, func_type, func_value)\ + __PYX__VERIFY_RETURN_INT(target_type, func_type, func_value, 1) +#define __PYX__VERIFY_RETURN_INT(target_type, func_type, func_value, exc)\ + {\ + func_type value = func_value;\ + if (sizeof(target_type) < sizeof(func_type)) {\ + if (unlikely(value != (func_type) (target_type) value)) {\ + func_type zero = 0;\ + if (exc && unlikely(value == (func_type)-1 && PyErr_Occurred()))\ + return (target_type) -1;\ + if (is_unsigned && unlikely(value < zero))\ + goto raise_neg_overflow;\ + else\ + goto raise_overflow;\ + }\ + }\ + return (target_type) value;\ + } + +/* Declarations */ + #if CYTHON_CCOMPLEX + #ifdef __cplusplus + static CYTHON_INLINE __pyx_t_float_complex __pyx_t_float_complex_from_parts(float x, float y) { + return ::std::complex< float >(x, y); + } + #else + static CYTHON_INLINE __pyx_t_float_complex __pyx_t_float_complex_from_parts(float x, float y) { + return x + y*(__pyx_t_float_complex)_Complex_I; + } + #endif +#else + static CYTHON_INLINE __pyx_t_float_complex __pyx_t_float_complex_from_parts(float x, float y) { + __pyx_t_float_complex z; + z.real = x; + z.imag = y; + return z; + } +#endif + +/* Arithmetic */ + #if CYTHON_CCOMPLEX +#else + static CYTHON_INLINE int __Pyx_c_eq_float(__pyx_t_float_complex a, __pyx_t_float_complex b) { + return (a.real == b.real) && (a.imag == b.imag); + } + static CYTHON_INLINE __pyx_t_float_complex __Pyx_c_sum_float(__pyx_t_float_complex a, __pyx_t_float_complex b) { + __pyx_t_float_complex z; + z.real = a.real + b.real; + z.imag = a.imag + b.imag; + return z; + } + static CYTHON_INLINE __pyx_t_float_complex __Pyx_c_diff_float(__pyx_t_float_complex a, __pyx_t_float_complex b) { + __pyx_t_float_complex z; + z.real = a.real - b.real; + z.imag = a.imag - b.imag; + return z; + } + static CYTHON_INLINE __pyx_t_float_complex __Pyx_c_prod_float(__pyx_t_float_complex a, __pyx_t_float_complex b) { + __pyx_t_float_complex z; + z.real = a.real * b.real - a.imag * b.imag; + z.imag = a.real * b.imag + a.imag * b.real; + return z; + } + #if 1 + static CYTHON_INLINE __pyx_t_float_complex __Pyx_c_quot_float(__pyx_t_float_complex a, __pyx_t_float_complex b) { + if (b.imag == 0) { + return __pyx_t_float_complex_from_parts(a.real / b.real, a.imag / b.real); + } else if (fabsf(b.real) >= fabsf(b.imag)) { + if (b.real == 0 && b.imag == 0) { + return __pyx_t_float_complex_from_parts(a.real / b.real, a.imag / b.imag); + } else { + float r = b.imag / b.real; + float s = (float)(1.0) / (b.real + b.imag * r); + return __pyx_t_float_complex_from_parts( + (a.real + a.imag * r) * s, (a.imag - a.real * r) * s); + } + } else { + float r = b.real / b.imag; + float s = (float)(1.0) / (b.imag + b.real * r); + return __pyx_t_float_complex_from_parts( + (a.real * r + a.imag) * s, (a.imag * r - a.real) * s); + } + } + #else + static CYTHON_INLINE __pyx_t_float_complex __Pyx_c_quot_float(__pyx_t_float_complex a, __pyx_t_float_complex b) { + if (b.imag == 0) { + return __pyx_t_float_complex_from_parts(a.real / b.real, a.imag / b.real); + } else { + float denom = b.real * b.real + b.imag * b.imag; + return __pyx_t_float_complex_from_parts( + (a.real * b.real + a.imag * b.imag) / denom, + (a.imag * b.real - a.real * b.imag) / denom); + } + } + #endif + static CYTHON_INLINE __pyx_t_float_complex __Pyx_c_neg_float(__pyx_t_float_complex a) { + __pyx_t_float_complex z; + z.real = -a.real; + z.imag = -a.imag; + return z; + } + static CYTHON_INLINE int __Pyx_c_is_zero_float(__pyx_t_float_complex a) { + return (a.real == 0) && (a.imag == 0); + } + static CYTHON_INLINE __pyx_t_float_complex __Pyx_c_conj_float(__pyx_t_float_complex a) { + __pyx_t_float_complex z; + z.real = a.real; + z.imag = -a.imag; + return z; + } + #if 1 + static CYTHON_INLINE float __Pyx_c_abs_float(__pyx_t_float_complex z) { + #if !defined(HAVE_HYPOT) || defined(_MSC_VER) + return sqrtf(z.real*z.real + z.imag*z.imag); + #else + return hypotf(z.real, z.imag); + #endif + } + static CYTHON_INLINE __pyx_t_float_complex __Pyx_c_pow_float(__pyx_t_float_complex a, __pyx_t_float_complex b) { + __pyx_t_float_complex z; + float r, lnr, theta, z_r, z_theta; + if (b.imag == 0 && b.real == (int)b.real) { + if (b.real < 0) { + float denom = a.real * a.real + a.imag * a.imag; + a.real = a.real / denom; + a.imag = -a.imag / denom; + b.real = -b.real; + } + switch ((int)b.real) { + case 0: + z.real = 1; + z.imag = 0; + return z; + case 1: + return a; + case 2: + return __Pyx_c_prod_float(a, a); + case 3: + z = __Pyx_c_prod_float(a, a); + return __Pyx_c_prod_float(z, a); + case 4: + z = __Pyx_c_prod_float(a, a); + return __Pyx_c_prod_float(z, z); + } + } + if (a.imag == 0) { + if (a.real == 0) { + return a; + } else if ((b.imag == 0) && (a.real >= 0)) { + z.real = powf(a.real, b.real); + z.imag = 0; + return z; + } else if (a.real > 0) { + r = a.real; + theta = 0; + } else { + r = -a.real; + theta = atan2f(0.0, -1.0); + } + } else { + r = __Pyx_c_abs_float(a); + theta = atan2f(a.imag, a.real); + } + lnr = logf(r); + z_r = expf(lnr * b.real - theta * b.imag); + z_theta = theta * b.real + lnr * b.imag; + z.real = z_r * cosf(z_theta); + z.imag = z_r * sinf(z_theta); + return z; + } + #endif +#endif + +/* Declarations */ + #if CYTHON_CCOMPLEX + #ifdef __cplusplus + static CYTHON_INLINE __pyx_t_double_complex __pyx_t_double_complex_from_parts(double x, double y) { + return ::std::complex< double >(x, y); + } + #else + static CYTHON_INLINE __pyx_t_double_complex __pyx_t_double_complex_from_parts(double x, double y) { + return x + y*(__pyx_t_double_complex)_Complex_I; + } + #endif +#else + static CYTHON_INLINE __pyx_t_double_complex __pyx_t_double_complex_from_parts(double x, double y) { + __pyx_t_double_complex z; + z.real = x; + z.imag = y; + return z; + } +#endif + +/* Arithmetic */ + #if CYTHON_CCOMPLEX +#else + static CYTHON_INLINE int __Pyx_c_eq_double(__pyx_t_double_complex a, __pyx_t_double_complex b) { + return (a.real == b.real) && (a.imag == b.imag); + } + static CYTHON_INLINE __pyx_t_double_complex __Pyx_c_sum_double(__pyx_t_double_complex a, __pyx_t_double_complex b) { + __pyx_t_double_complex z; + z.real = a.real + b.real; + z.imag = a.imag + b.imag; + return z; + } + static CYTHON_INLINE __pyx_t_double_complex __Pyx_c_diff_double(__pyx_t_double_complex a, __pyx_t_double_complex b) { + __pyx_t_double_complex z; + z.real = a.real - b.real; + z.imag = a.imag - b.imag; + return z; + } + static CYTHON_INLINE __pyx_t_double_complex __Pyx_c_prod_double(__pyx_t_double_complex a, __pyx_t_double_complex b) { + __pyx_t_double_complex z; + z.real = a.real * b.real - a.imag * b.imag; + z.imag = a.real * b.imag + a.imag * b.real; + return z; + } + #if 1 + static CYTHON_INLINE __pyx_t_double_complex __Pyx_c_quot_double(__pyx_t_double_complex a, __pyx_t_double_complex b) { + if (b.imag == 0) { + return __pyx_t_double_complex_from_parts(a.real / b.real, a.imag / b.real); + } else if (fabs(b.real) >= fabs(b.imag)) { + if (b.real == 0 && b.imag == 0) { + return __pyx_t_double_complex_from_parts(a.real / b.real, a.imag / b.imag); + } else { + double r = b.imag / b.real; + double s = (double)(1.0) / (b.real + b.imag * r); + return __pyx_t_double_complex_from_parts( + (a.real + a.imag * r) * s, (a.imag - a.real * r) * s); + } + } else { + double r = b.real / b.imag; + double s = (double)(1.0) / (b.imag + b.real * r); + return __pyx_t_double_complex_from_parts( + (a.real * r + a.imag) * s, (a.imag * r - a.real) * s); + } + } + #else + static CYTHON_INLINE __pyx_t_double_complex __Pyx_c_quot_double(__pyx_t_double_complex a, __pyx_t_double_complex b) { + if (b.imag == 0) { + return __pyx_t_double_complex_from_parts(a.real / b.real, a.imag / b.real); + } else { + double denom = b.real * b.real + b.imag * b.imag; + return __pyx_t_double_complex_from_parts( + (a.real * b.real + a.imag * b.imag) / denom, + (a.imag * b.real - a.real * b.imag) / denom); + } + } + #endif + static CYTHON_INLINE __pyx_t_double_complex __Pyx_c_neg_double(__pyx_t_double_complex a) { + __pyx_t_double_complex z; + z.real = -a.real; + z.imag = -a.imag; + return z; + } + static CYTHON_INLINE int __Pyx_c_is_zero_double(__pyx_t_double_complex a) { + return (a.real == 0) && (a.imag == 0); + } + static CYTHON_INLINE __pyx_t_double_complex __Pyx_c_conj_double(__pyx_t_double_complex a) { + __pyx_t_double_complex z; + z.real = a.real; + z.imag = -a.imag; + return z; + } + #if 1 + static CYTHON_INLINE double __Pyx_c_abs_double(__pyx_t_double_complex z) { + #if !defined(HAVE_HYPOT) || defined(_MSC_VER) + return sqrt(z.real*z.real + z.imag*z.imag); + #else + return hypot(z.real, z.imag); + #endif + } + static CYTHON_INLINE __pyx_t_double_complex __Pyx_c_pow_double(__pyx_t_double_complex a, __pyx_t_double_complex b) { + __pyx_t_double_complex z; + double r, lnr, theta, z_r, z_theta; + if (b.imag == 0 && b.real == (int)b.real) { + if (b.real < 0) { + double denom = a.real * a.real + a.imag * a.imag; + a.real = a.real / denom; + a.imag = -a.imag / denom; + b.real = -b.real; + } + switch ((int)b.real) { + case 0: + z.real = 1; + z.imag = 0; + return z; + case 1: + return a; + case 2: + return __Pyx_c_prod_double(a, a); + case 3: + z = __Pyx_c_prod_double(a, a); + return __Pyx_c_prod_double(z, a); + case 4: + z = __Pyx_c_prod_double(a, a); + return __Pyx_c_prod_double(z, z); + } + } + if (a.imag == 0) { + if (a.real == 0) { + return a; + } else if ((b.imag == 0) && (a.real >= 0)) { + z.real = pow(a.real, b.real); + z.imag = 0; + return z; + } else if (a.real > 0) { + r = a.real; + theta = 0; + } else { + r = -a.real; + theta = atan2(0.0, -1.0); + } + } else { + r = __Pyx_c_abs_double(a); + theta = atan2(a.imag, a.real); + } + lnr = log(r); + z_r = exp(lnr * b.real - theta * b.imag); + z_theta = theta * b.real + lnr * b.imag; + z.real = z_r * cos(z_theta); + z.imag = z_r * sin(z_theta); + return z; + } + #endif +#endif + +/* CIntToPy */ + static CYTHON_INLINE PyObject* __Pyx_PyInt_From_unsigned_int(unsigned int value) { +#ifdef __Pyx_HAS_GCC_DIAGNOSTIC +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wconversion" +#endif + const unsigned int neg_one = (unsigned int) -1, const_zero = (unsigned int) 0; +#ifdef __Pyx_HAS_GCC_DIAGNOSTIC +#pragma GCC diagnostic pop +#endif + const int is_unsigned = neg_one > const_zero; + if (is_unsigned) { + if (sizeof(unsigned int) < sizeof(long)) { + return PyInt_FromLong((long) value); + } else if (sizeof(unsigned int) <= sizeof(unsigned long)) { + return PyLong_FromUnsignedLong((unsigned long) value); +#ifdef HAVE_LONG_LONG + } else if (sizeof(unsigned int) <= sizeof(unsigned PY_LONG_LONG)) { + return PyLong_FromUnsignedLongLong((unsigned PY_LONG_LONG) value); +#endif + } + } else { + if (sizeof(unsigned int) <= sizeof(long)) { + return PyInt_FromLong((long) value); +#ifdef HAVE_LONG_LONG + } else if (sizeof(unsigned int) <= sizeof(PY_LONG_LONG)) { + return PyLong_FromLongLong((PY_LONG_LONG) value); +#endif + } + } + { + int one = 1; int little = (int)*(unsigned char *)&one; + unsigned char *bytes = (unsigned char *)&value; + return _PyLong_FromByteArray(bytes, sizeof(unsigned int), + little, !is_unsigned); + } +} + +/* CIntFromPy */ + static CYTHON_INLINE unsigned int __Pyx_PyInt_As_unsigned_int(PyObject *x) { +#ifdef __Pyx_HAS_GCC_DIAGNOSTIC +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wconversion" +#endif + const unsigned int neg_one = (unsigned int) -1, const_zero = (unsigned int) 0; +#ifdef __Pyx_HAS_GCC_DIAGNOSTIC +#pragma GCC diagnostic pop +#endif + const int is_unsigned = neg_one > const_zero; +#if PY_MAJOR_VERSION < 3 + if (likely(PyInt_Check(x))) { + if (sizeof(unsigned int) < sizeof(long)) { + __PYX_VERIFY_RETURN_INT(unsigned int, long, PyInt_AS_LONG(x)) + } else { + long val = PyInt_AS_LONG(x); + if (is_unsigned && unlikely(val < 0)) { + goto raise_neg_overflow; + } + return (unsigned int) val; + } + } else +#endif + if (likely(PyLong_Check(x))) { + if (is_unsigned) { +#if CYTHON_USE_PYLONG_INTERNALS + const digit* digits = ((PyLongObject*)x)->ob_digit; + switch (Py_SIZE(x)) { + case 0: return (unsigned int) 0; + case 1: __PYX_VERIFY_RETURN_INT(unsigned int, digit, digits[0]) + case 2: + if (8 * sizeof(unsigned int) > 1 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 2 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(unsigned int, unsigned long, (((((unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(unsigned int) >= 2 * PyLong_SHIFT) { + return (unsigned int) (((((unsigned int)digits[1]) << PyLong_SHIFT) | (unsigned int)digits[0])); + } + } + break; + case 3: + if (8 * sizeof(unsigned int) > 2 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 3 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(unsigned int, unsigned long, (((((((unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(unsigned int) >= 3 * PyLong_SHIFT) { + return (unsigned int) (((((((unsigned int)digits[2]) << PyLong_SHIFT) | (unsigned int)digits[1]) << PyLong_SHIFT) | (unsigned int)digits[0])); + } + } + break; + case 4: + if (8 * sizeof(unsigned int) > 3 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 4 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(unsigned int, unsigned long, (((((((((unsigned long)digits[3]) << PyLong_SHIFT) | (unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(unsigned int) >= 4 * PyLong_SHIFT) { + return (unsigned int) (((((((((unsigned int)digits[3]) << PyLong_SHIFT) | (unsigned int)digits[2]) << PyLong_SHIFT) | (unsigned int)digits[1]) << PyLong_SHIFT) | (unsigned int)digits[0])); + } + } + break; + } +#endif +#if CYTHON_COMPILING_IN_CPYTHON && PY_VERSION_HEX < 0x030C00A7 + if (unlikely(Py_SIZE(x) < 0)) { + goto raise_neg_overflow; + } +#else + { + int result = PyObject_RichCompareBool(x, Py_False, Py_LT); + if (unlikely(result < 0)) + return (unsigned int) -1; + if (unlikely(result == 1)) + goto raise_neg_overflow; + } +#endif + if (sizeof(unsigned int) <= sizeof(unsigned long)) { + __PYX_VERIFY_RETURN_INT_EXC(unsigned int, unsigned long, PyLong_AsUnsignedLong(x)) +#ifdef HAVE_LONG_LONG + } else if (sizeof(unsigned int) <= sizeof(unsigned PY_LONG_LONG)) { + __PYX_VERIFY_RETURN_INT_EXC(unsigned int, unsigned PY_LONG_LONG, PyLong_AsUnsignedLongLong(x)) +#endif + } + } else { +#if CYTHON_USE_PYLONG_INTERNALS + const digit* digits = ((PyLongObject*)x)->ob_digit; + switch (Py_SIZE(x)) { + case 0: return (unsigned int) 0; + case -1: __PYX_VERIFY_RETURN_INT(unsigned int, sdigit, (sdigit) (-(sdigit)digits[0])) + case 1: __PYX_VERIFY_RETURN_INT(unsigned int, digit, +digits[0]) + case -2: + if (8 * sizeof(unsigned int) - 1 > 1 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 2 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(unsigned int, long, -(long) (((((unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(unsigned int) - 1 > 2 * PyLong_SHIFT) { + return (unsigned int) (((unsigned int)-1)*(((((unsigned int)digits[1]) << PyLong_SHIFT) | (unsigned int)digits[0]))); + } + } + break; + case 2: + if (8 * sizeof(unsigned int) > 1 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 2 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(unsigned int, unsigned long, (((((unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(unsigned int) - 1 > 2 * PyLong_SHIFT) { + return (unsigned int) ((((((unsigned int)digits[1]) << PyLong_SHIFT) | (unsigned int)digits[0]))); + } + } + break; + case -3: + if (8 * sizeof(unsigned int) - 1 > 2 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 3 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(unsigned int, long, -(long) (((((((unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(unsigned int) - 1 > 3 * PyLong_SHIFT) { + return (unsigned int) (((unsigned int)-1)*(((((((unsigned int)digits[2]) << PyLong_SHIFT) | (unsigned int)digits[1]) << PyLong_SHIFT) | (unsigned int)digits[0]))); + } + } + break; + case 3: + if (8 * sizeof(unsigned int) > 2 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 3 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(unsigned int, unsigned long, (((((((unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(unsigned int) - 1 > 3 * PyLong_SHIFT) { + return (unsigned int) ((((((((unsigned int)digits[2]) << PyLong_SHIFT) | (unsigned int)digits[1]) << PyLong_SHIFT) | (unsigned int)digits[0]))); + } + } + break; + case -4: + if (8 * sizeof(unsigned int) - 1 > 3 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 4 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(unsigned int, long, -(long) (((((((((unsigned long)digits[3]) << PyLong_SHIFT) | (unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(unsigned int) - 1 > 4 * PyLong_SHIFT) { + return (unsigned int) (((unsigned int)-1)*(((((((((unsigned int)digits[3]) << PyLong_SHIFT) | (unsigned int)digits[2]) << PyLong_SHIFT) | (unsigned int)digits[1]) << PyLong_SHIFT) | (unsigned int)digits[0]))); + } + } + break; + case 4: + if (8 * sizeof(unsigned int) > 3 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 4 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(unsigned int, unsigned long, (((((((((unsigned long)digits[3]) << PyLong_SHIFT) | (unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(unsigned int) - 1 > 4 * PyLong_SHIFT) { + return (unsigned int) ((((((((((unsigned int)digits[3]) << PyLong_SHIFT) | (unsigned int)digits[2]) << PyLong_SHIFT) | (unsigned int)digits[1]) << PyLong_SHIFT) | (unsigned int)digits[0]))); + } + } + break; + } +#endif + if (sizeof(unsigned int) <= sizeof(long)) { + __PYX_VERIFY_RETURN_INT_EXC(unsigned int, long, PyLong_AsLong(x)) +#ifdef HAVE_LONG_LONG + } else if (sizeof(unsigned int) <= sizeof(PY_LONG_LONG)) { + __PYX_VERIFY_RETURN_INT_EXC(unsigned int, PY_LONG_LONG, PyLong_AsLongLong(x)) +#endif + } + } + { +#if CYTHON_COMPILING_IN_PYPY && !defined(_PyLong_AsByteArray) + PyErr_SetString(PyExc_RuntimeError, + "_PyLong_AsByteArray() not available in PyPy, cannot convert large numbers"); +#else + unsigned int val; + PyObject *v = __Pyx_PyNumber_IntOrLong(x); + #if PY_MAJOR_VERSION < 3 + if (likely(v) && !PyLong_Check(v)) { + PyObject *tmp = v; + v = PyNumber_Long(tmp); + Py_DECREF(tmp); + } + #endif + if (likely(v)) { + int one = 1; int is_little = (int)*(unsigned char *)&one; + unsigned char *bytes = (unsigned char *)&val; + int ret = _PyLong_AsByteArray((PyLongObject *)v, + bytes, sizeof(val), + is_little, !is_unsigned); + Py_DECREF(v); + if (likely(!ret)) + return val; + } +#endif + return (unsigned int) -1; + } + } else { + unsigned int val; + PyObject *tmp = __Pyx_PyNumber_IntOrLong(x); + if (!tmp) return (unsigned int) -1; + val = __Pyx_PyInt_As_unsigned_int(tmp); + Py_DECREF(tmp); + return val; + } +raise_overflow: + PyErr_SetString(PyExc_OverflowError, + "value too large to convert to unsigned int"); + return (unsigned int) -1; +raise_neg_overflow: + PyErr_SetString(PyExc_OverflowError, + "can't convert negative value to unsigned int"); + return (unsigned int) -1; +} + +/* CIntToPy */ + static CYTHON_INLINE PyObject* __Pyx_PyInt_From_long(long value) { +#ifdef __Pyx_HAS_GCC_DIAGNOSTIC +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wconversion" +#endif + const long neg_one = (long) -1, const_zero = (long) 0; +#ifdef __Pyx_HAS_GCC_DIAGNOSTIC +#pragma GCC diagnostic pop +#endif + const int is_unsigned = neg_one > const_zero; + if (is_unsigned) { + if (sizeof(long) < sizeof(long)) { + return PyInt_FromLong((long) value); + } else if (sizeof(long) <= sizeof(unsigned long)) { + return PyLong_FromUnsignedLong((unsigned long) value); +#ifdef HAVE_LONG_LONG + } else if (sizeof(long) <= sizeof(unsigned PY_LONG_LONG)) { + return PyLong_FromUnsignedLongLong((unsigned PY_LONG_LONG) value); +#endif + } + } else { + if (sizeof(long) <= sizeof(long)) { + return PyInt_FromLong((long) value); +#ifdef HAVE_LONG_LONG + } else if (sizeof(long) <= sizeof(PY_LONG_LONG)) { + return PyLong_FromLongLong((PY_LONG_LONG) value); +#endif + } + } + { + int one = 1; int little = (int)*(unsigned char *)&one; + unsigned char *bytes = (unsigned char *)&value; + return _PyLong_FromByteArray(bytes, sizeof(long), + little, !is_unsigned); + } +} + +/* CIntFromPy */ + static CYTHON_INLINE long __Pyx_PyInt_As_long(PyObject *x) { +#ifdef __Pyx_HAS_GCC_DIAGNOSTIC +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wconversion" +#endif + const long neg_one = (long) -1, const_zero = (long) 0; +#ifdef __Pyx_HAS_GCC_DIAGNOSTIC +#pragma GCC diagnostic pop +#endif + const int is_unsigned = neg_one > const_zero; +#if PY_MAJOR_VERSION < 3 + if (likely(PyInt_Check(x))) { + if (sizeof(long) < sizeof(long)) { + __PYX_VERIFY_RETURN_INT(long, long, PyInt_AS_LONG(x)) + } else { + long val = PyInt_AS_LONG(x); + if (is_unsigned && unlikely(val < 0)) { + goto raise_neg_overflow; + } + return (long) val; + } + } else +#endif + if (likely(PyLong_Check(x))) { + if (is_unsigned) { +#if CYTHON_USE_PYLONG_INTERNALS + const digit* digits = ((PyLongObject*)x)->ob_digit; + switch (Py_SIZE(x)) { + case 0: return (long) 0; + case 1: __PYX_VERIFY_RETURN_INT(long, digit, digits[0]) + case 2: + if (8 * sizeof(long) > 1 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 2 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(long, unsigned long, (((((unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(long) >= 2 * PyLong_SHIFT) { + return (long) (((((long)digits[1]) << PyLong_SHIFT) | (long)digits[0])); + } + } + break; + case 3: + if (8 * sizeof(long) > 2 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 3 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(long, unsigned long, (((((((unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(long) >= 3 * PyLong_SHIFT) { + return (long) (((((((long)digits[2]) << PyLong_SHIFT) | (long)digits[1]) << PyLong_SHIFT) | (long)digits[0])); + } + } + break; + case 4: + if (8 * sizeof(long) > 3 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 4 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(long, unsigned long, (((((((((unsigned long)digits[3]) << PyLong_SHIFT) | (unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(long) >= 4 * PyLong_SHIFT) { + return (long) (((((((((long)digits[3]) << PyLong_SHIFT) | (long)digits[2]) << PyLong_SHIFT) | (long)digits[1]) << PyLong_SHIFT) | (long)digits[0])); + } + } + break; + } +#endif +#if CYTHON_COMPILING_IN_CPYTHON && PY_VERSION_HEX < 0x030C00A7 + if (unlikely(Py_SIZE(x) < 0)) { + goto raise_neg_overflow; + } +#else + { + int result = PyObject_RichCompareBool(x, Py_False, Py_LT); + if (unlikely(result < 0)) + return (long) -1; + if (unlikely(result == 1)) + goto raise_neg_overflow; + } +#endif + if (sizeof(long) <= sizeof(unsigned long)) { + __PYX_VERIFY_RETURN_INT_EXC(long, unsigned long, PyLong_AsUnsignedLong(x)) +#ifdef HAVE_LONG_LONG + } else if (sizeof(long) <= sizeof(unsigned PY_LONG_LONG)) { + __PYX_VERIFY_RETURN_INT_EXC(long, unsigned PY_LONG_LONG, PyLong_AsUnsignedLongLong(x)) +#endif + } + } else { +#if CYTHON_USE_PYLONG_INTERNALS + const digit* digits = ((PyLongObject*)x)->ob_digit; + switch (Py_SIZE(x)) { + case 0: return (long) 0; + case -1: __PYX_VERIFY_RETURN_INT(long, sdigit, (sdigit) (-(sdigit)digits[0])) + case 1: __PYX_VERIFY_RETURN_INT(long, digit, +digits[0]) + case -2: + if (8 * sizeof(long) - 1 > 1 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 2 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(long, long, -(long) (((((unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(long) - 1 > 2 * PyLong_SHIFT) { + return (long) (((long)-1)*(((((long)digits[1]) << PyLong_SHIFT) | (long)digits[0]))); + } + } + break; + case 2: + if (8 * sizeof(long) > 1 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 2 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(long, unsigned long, (((((unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(long) - 1 > 2 * PyLong_SHIFT) { + return (long) ((((((long)digits[1]) << PyLong_SHIFT) | (long)digits[0]))); + } + } + break; + case -3: + if (8 * sizeof(long) - 1 > 2 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 3 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(long, long, -(long) (((((((unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(long) - 1 > 3 * PyLong_SHIFT) { + return (long) (((long)-1)*(((((((long)digits[2]) << PyLong_SHIFT) | (long)digits[1]) << PyLong_SHIFT) | (long)digits[0]))); + } + } + break; + case 3: + if (8 * sizeof(long) > 2 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 3 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(long, unsigned long, (((((((unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(long) - 1 > 3 * PyLong_SHIFT) { + return (long) ((((((((long)digits[2]) << PyLong_SHIFT) | (long)digits[1]) << PyLong_SHIFT) | (long)digits[0]))); + } + } + break; + case -4: + if (8 * sizeof(long) - 1 > 3 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 4 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(long, long, -(long) (((((((((unsigned long)digits[3]) << PyLong_SHIFT) | (unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(long) - 1 > 4 * PyLong_SHIFT) { + return (long) (((long)-1)*(((((((((long)digits[3]) << PyLong_SHIFT) | (long)digits[2]) << PyLong_SHIFT) | (long)digits[1]) << PyLong_SHIFT) | (long)digits[0]))); + } + } + break; + case 4: + if (8 * sizeof(long) > 3 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 4 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(long, unsigned long, (((((((((unsigned long)digits[3]) << PyLong_SHIFT) | (unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(long) - 1 > 4 * PyLong_SHIFT) { + return (long) ((((((((((long)digits[3]) << PyLong_SHIFT) | (long)digits[2]) << PyLong_SHIFT) | (long)digits[1]) << PyLong_SHIFT) | (long)digits[0]))); + } + } + break; + } +#endif + if (sizeof(long) <= sizeof(long)) { + __PYX_VERIFY_RETURN_INT_EXC(long, long, PyLong_AsLong(x)) +#ifdef HAVE_LONG_LONG + } else if (sizeof(long) <= sizeof(PY_LONG_LONG)) { + __PYX_VERIFY_RETURN_INT_EXC(long, PY_LONG_LONG, PyLong_AsLongLong(x)) +#endif + } + } + { +#if CYTHON_COMPILING_IN_PYPY && !defined(_PyLong_AsByteArray) + PyErr_SetString(PyExc_RuntimeError, + "_PyLong_AsByteArray() not available in PyPy, cannot convert large numbers"); +#else + long val; + PyObject *v = __Pyx_PyNumber_IntOrLong(x); + #if PY_MAJOR_VERSION < 3 + if (likely(v) && !PyLong_Check(v)) { + PyObject *tmp = v; + v = PyNumber_Long(tmp); + Py_DECREF(tmp); + } + #endif + if (likely(v)) { + int one = 1; int is_little = (int)*(unsigned char *)&one; + unsigned char *bytes = (unsigned char *)&val; + int ret = _PyLong_AsByteArray((PyLongObject *)v, + bytes, sizeof(val), + is_little, !is_unsigned); + Py_DECREF(v); + if (likely(!ret)) + return val; + } +#endif + return (long) -1; + } + } else { + long val; + PyObject *tmp = __Pyx_PyNumber_IntOrLong(x); + if (!tmp) return (long) -1; + val = __Pyx_PyInt_As_long(tmp); + Py_DECREF(tmp); + return val; + } +raise_overflow: + PyErr_SetString(PyExc_OverflowError, + "value too large to convert to long"); + return (long) -1; +raise_neg_overflow: + PyErr_SetString(PyExc_OverflowError, + "can't convert negative value to long"); + return (long) -1; +} + +/* CIntFromPy */ + static CYTHON_INLINE int __Pyx_PyInt_As_int(PyObject *x) { +#ifdef __Pyx_HAS_GCC_DIAGNOSTIC +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wconversion" +#endif + const int neg_one = (int) -1, const_zero = (int) 0; +#ifdef __Pyx_HAS_GCC_DIAGNOSTIC +#pragma GCC diagnostic pop +#endif + const int is_unsigned = neg_one > const_zero; +#if PY_MAJOR_VERSION < 3 + if (likely(PyInt_Check(x))) { + if (sizeof(int) < sizeof(long)) { + __PYX_VERIFY_RETURN_INT(int, long, PyInt_AS_LONG(x)) + } else { + long val = PyInt_AS_LONG(x); + if (is_unsigned && unlikely(val < 0)) { + goto raise_neg_overflow; + } + return (int) val; + } + } else +#endif + if (likely(PyLong_Check(x))) { + if (is_unsigned) { +#if CYTHON_USE_PYLONG_INTERNALS + const digit* digits = ((PyLongObject*)x)->ob_digit; + switch (Py_SIZE(x)) { + case 0: return (int) 0; + case 1: __PYX_VERIFY_RETURN_INT(int, digit, digits[0]) + case 2: + if (8 * sizeof(int) > 1 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 2 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(int, unsigned long, (((((unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(int) >= 2 * PyLong_SHIFT) { + return (int) (((((int)digits[1]) << PyLong_SHIFT) | (int)digits[0])); + } + } + break; + case 3: + if (8 * sizeof(int) > 2 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 3 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(int, unsigned long, (((((((unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(int) >= 3 * PyLong_SHIFT) { + return (int) (((((((int)digits[2]) << PyLong_SHIFT) | (int)digits[1]) << PyLong_SHIFT) | (int)digits[0])); + } + } + break; + case 4: + if (8 * sizeof(int) > 3 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 4 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(int, unsigned long, (((((((((unsigned long)digits[3]) << PyLong_SHIFT) | (unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(int) >= 4 * PyLong_SHIFT) { + return (int) (((((((((int)digits[3]) << PyLong_SHIFT) | (int)digits[2]) << PyLong_SHIFT) | (int)digits[1]) << PyLong_SHIFT) | (int)digits[0])); + } + } + break; + } +#endif +#if CYTHON_COMPILING_IN_CPYTHON && PY_VERSION_HEX < 0x030C00A7 + if (unlikely(Py_SIZE(x) < 0)) { + goto raise_neg_overflow; + } +#else + { + int result = PyObject_RichCompareBool(x, Py_False, Py_LT); + if (unlikely(result < 0)) + return (int) -1; + if (unlikely(result == 1)) + goto raise_neg_overflow; + } +#endif + if (sizeof(int) <= sizeof(unsigned long)) { + __PYX_VERIFY_RETURN_INT_EXC(int, unsigned long, PyLong_AsUnsignedLong(x)) +#ifdef HAVE_LONG_LONG + } else if (sizeof(int) <= sizeof(unsigned PY_LONG_LONG)) { + __PYX_VERIFY_RETURN_INT_EXC(int, unsigned PY_LONG_LONG, PyLong_AsUnsignedLongLong(x)) +#endif + } + } else { +#if CYTHON_USE_PYLONG_INTERNALS + const digit* digits = ((PyLongObject*)x)->ob_digit; + switch (Py_SIZE(x)) { + case 0: return (int) 0; + case -1: __PYX_VERIFY_RETURN_INT(int, sdigit, (sdigit) (-(sdigit)digits[0])) + case 1: __PYX_VERIFY_RETURN_INT(int, digit, +digits[0]) + case -2: + if (8 * sizeof(int) - 1 > 1 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 2 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(int, long, -(long) (((((unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(int) - 1 > 2 * PyLong_SHIFT) { + return (int) (((int)-1)*(((((int)digits[1]) << PyLong_SHIFT) | (int)digits[0]))); + } + } + break; + case 2: + if (8 * sizeof(int) > 1 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 2 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(int, unsigned long, (((((unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(int) - 1 > 2 * PyLong_SHIFT) { + return (int) ((((((int)digits[1]) << PyLong_SHIFT) | (int)digits[0]))); + } + } + break; + case -3: + if (8 * sizeof(int) - 1 > 2 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 3 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(int, long, -(long) (((((((unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(int) - 1 > 3 * PyLong_SHIFT) { + return (int) (((int)-1)*(((((((int)digits[2]) << PyLong_SHIFT) | (int)digits[1]) << PyLong_SHIFT) | (int)digits[0]))); + } + } + break; + case 3: + if (8 * sizeof(int) > 2 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 3 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(int, unsigned long, (((((((unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(int) - 1 > 3 * PyLong_SHIFT) { + return (int) ((((((((int)digits[2]) << PyLong_SHIFT) | (int)digits[1]) << PyLong_SHIFT) | (int)digits[0]))); + } + } + break; + case -4: + if (8 * sizeof(int) - 1 > 3 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 4 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(int, long, -(long) (((((((((unsigned long)digits[3]) << PyLong_SHIFT) | (unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(int) - 1 > 4 * PyLong_SHIFT) { + return (int) (((int)-1)*(((((((((int)digits[3]) << PyLong_SHIFT) | (int)digits[2]) << PyLong_SHIFT) | (int)digits[1]) << PyLong_SHIFT) | (int)digits[0]))); + } + } + break; + case 4: + if (8 * sizeof(int) > 3 * PyLong_SHIFT) { + if (8 * sizeof(unsigned long) > 4 * PyLong_SHIFT) { + __PYX_VERIFY_RETURN_INT(int, unsigned long, (((((((((unsigned long)digits[3]) << PyLong_SHIFT) | (unsigned long)digits[2]) << PyLong_SHIFT) | (unsigned long)digits[1]) << PyLong_SHIFT) | (unsigned long)digits[0]))) + } else if (8 * sizeof(int) - 1 > 4 * PyLong_SHIFT) { + return (int) ((((((((((int)digits[3]) << PyLong_SHIFT) | (int)digits[2]) << PyLong_SHIFT) | (int)digits[1]) << PyLong_SHIFT) | (int)digits[0]))); + } + } + break; + } +#endif + if (sizeof(int) <= sizeof(long)) { + __PYX_VERIFY_RETURN_INT_EXC(int, long, PyLong_AsLong(x)) +#ifdef HAVE_LONG_LONG + } else if (sizeof(int) <= sizeof(PY_LONG_LONG)) { + __PYX_VERIFY_RETURN_INT_EXC(int, PY_LONG_LONG, PyLong_AsLongLong(x)) +#endif + } + } + { +#if CYTHON_COMPILING_IN_PYPY && !defined(_PyLong_AsByteArray) + PyErr_SetString(PyExc_RuntimeError, + "_PyLong_AsByteArray() not available in PyPy, cannot convert large numbers"); +#else + int val; + PyObject *v = __Pyx_PyNumber_IntOrLong(x); + #if PY_MAJOR_VERSION < 3 + if (likely(v) && !PyLong_Check(v)) { + PyObject *tmp = v; + v = PyNumber_Long(tmp); + Py_DECREF(tmp); + } + #endif + if (likely(v)) { + int one = 1; int is_little = (int)*(unsigned char *)&one; + unsigned char *bytes = (unsigned char *)&val; + int ret = _PyLong_AsByteArray((PyLongObject *)v, + bytes, sizeof(val), + is_little, !is_unsigned); + Py_DECREF(v); + if (likely(!ret)) + return val; + } +#endif + return (int) -1; + } + } else { + int val; + PyObject *tmp = __Pyx_PyNumber_IntOrLong(x); + if (!tmp) return (int) -1; + val = __Pyx_PyInt_As_int(tmp); + Py_DECREF(tmp); + return val; + } +raise_overflow: + PyErr_SetString(PyExc_OverflowError, + "value too large to convert to int"); + return (int) -1; +raise_neg_overflow: + PyErr_SetString(PyExc_OverflowError, + "can't convert negative value to int"); + return (int) -1; +} + +/* FastTypeChecks */ + #if CYTHON_COMPILING_IN_CPYTHON +static int __Pyx_InBases(PyTypeObject *a, PyTypeObject *b) { + while (a) { + a = a->tp_base; + if (a == b) + return 1; + } + return b == &PyBaseObject_Type; +} +static CYTHON_INLINE int __Pyx_IsSubtype(PyTypeObject *a, PyTypeObject *b) { + PyObject *mro; + if (a == b) return 1; + mro = a->tp_mro; + if (likely(mro)) { + Py_ssize_t i, n; + n = PyTuple_GET_SIZE(mro); + for (i = 0; i < n; i++) { + if (PyTuple_GET_ITEM(mro, i) == (PyObject *)b) + return 1; + } + return 0; + } + return __Pyx_InBases(a, b); +} +#if PY_MAJOR_VERSION == 2 +static int __Pyx_inner_PyErr_GivenExceptionMatches2(PyObject *err, PyObject* exc_type1, PyObject* exc_type2) { + PyObject *exception, *value, *tb; + int res; + __Pyx_PyThreadState_declare + __Pyx_PyThreadState_assign + __Pyx_ErrFetch(&exception, &value, &tb); + res = exc_type1 ? PyObject_IsSubclass(err, exc_type1) : 0; + if (unlikely(res == -1)) { + PyErr_WriteUnraisable(err); + res = 0; + } + if (!res) { + res = PyObject_IsSubclass(err, exc_type2); + if (unlikely(res == -1)) { + PyErr_WriteUnraisable(err); + res = 0; + } + } + __Pyx_ErrRestore(exception, value, tb); + return res; +} +#else +static CYTHON_INLINE int __Pyx_inner_PyErr_GivenExceptionMatches2(PyObject *err, PyObject* exc_type1, PyObject *exc_type2) { + int res = exc_type1 ? __Pyx_IsSubtype((PyTypeObject*)err, (PyTypeObject*)exc_type1) : 0; + if (!res) { + res = __Pyx_IsSubtype((PyTypeObject*)err, (PyTypeObject*)exc_type2); + } + return res; +} +#endif +static int __Pyx_PyErr_GivenExceptionMatchesTuple(PyObject *exc_type, PyObject *tuple) { + Py_ssize_t i, n; + assert(PyExceptionClass_Check(exc_type)); + n = PyTuple_GET_SIZE(tuple); +#if PY_MAJOR_VERSION >= 3 + for (i=0; i '9'); + break; + } + if (rt_from_call[i] != ctversion[i]) { + same = 0; + break; + } + } + if (!same) { + char rtversion[5] = {'\0'}; + char message[200]; + for (i=0; i<4; ++i) { + if (rt_from_call[i] == '.') { + if (found_dot) break; + found_dot = 1; + } else if (rt_from_call[i] < '0' || rt_from_call[i] > '9') { + break; + } + rtversion[i] = rt_from_call[i]; + } + PyOS_snprintf(message, sizeof(message), + "compiletime version %s of module '%.100s' " + "does not match runtime version %s", + ctversion, __Pyx_MODULE_NAME, rtversion); + return PyErr_WarnEx(NULL, message, 1); + } + return 0; +} + +/* InitStrings */ + static int __Pyx_InitStrings(__Pyx_StringTabEntry *t) { + while (t->p) { + #if PY_MAJOR_VERSION < 3 + if (t->is_unicode) { + *t->p = PyUnicode_DecodeUTF8(t->s, t->n - 1, NULL); + } else if (t->intern) { + *t->p = PyString_InternFromString(t->s); + } else { + *t->p = PyString_FromStringAndSize(t->s, t->n - 1); + } + #else + if (t->is_unicode | t->is_str) { + if (t->intern) { + *t->p = PyUnicode_InternFromString(t->s); + } else if (t->encoding) { + *t->p = PyUnicode_Decode(t->s, t->n - 1, t->encoding, NULL); + } else { + *t->p = PyUnicode_FromStringAndSize(t->s, t->n - 1); + } + } else { + *t->p = PyBytes_FromStringAndSize(t->s, t->n - 1); + } + #endif + if (!*t->p) + return -1; + if (PyObject_Hash(*t->p) == -1) + return -1; + ++t; + } + return 0; +} + +static CYTHON_INLINE PyObject* __Pyx_PyUnicode_FromString(const char* c_str) { + return __Pyx_PyUnicode_FromStringAndSize(c_str, (Py_ssize_t)strlen(c_str)); +} +static CYTHON_INLINE const char* __Pyx_PyObject_AsString(PyObject* o) { + Py_ssize_t ignore; + return __Pyx_PyObject_AsStringAndSize(o, &ignore); +} +#if __PYX_DEFAULT_STRING_ENCODING_IS_ASCII || __PYX_DEFAULT_STRING_ENCODING_IS_DEFAULT +#if !CYTHON_PEP393_ENABLED +static const char* __Pyx_PyUnicode_AsStringAndSize(PyObject* o, Py_ssize_t *length) { + char* defenc_c; + PyObject* defenc = _PyUnicode_AsDefaultEncodedString(o, NULL); + if (!defenc) return NULL; + defenc_c = PyBytes_AS_STRING(defenc); +#if __PYX_DEFAULT_STRING_ENCODING_IS_ASCII + { + char* end = defenc_c + PyBytes_GET_SIZE(defenc); + char* c; + for (c = defenc_c; c < end; c++) { + if ((unsigned char) (*c) >= 128) { + PyUnicode_AsASCIIString(o); + return NULL; + } + } + } +#endif + *length = PyBytes_GET_SIZE(defenc); + return defenc_c; +} +#else +static CYTHON_INLINE const char* __Pyx_PyUnicode_AsStringAndSize(PyObject* o, Py_ssize_t *length) { + if (unlikely(__Pyx_PyUnicode_READY(o) == -1)) return NULL; +#if __PYX_DEFAULT_STRING_ENCODING_IS_ASCII + if (likely(PyUnicode_IS_ASCII(o))) { + *length = PyUnicode_GET_LENGTH(o); + return PyUnicode_AsUTF8(o); + } else { + PyUnicode_AsASCIIString(o); + return NULL; + } +#else + return PyUnicode_AsUTF8AndSize(o, length); +#endif +} +#endif +#endif +static CYTHON_INLINE const char* __Pyx_PyObject_AsStringAndSize(PyObject* o, Py_ssize_t *length) { +#if __PYX_DEFAULT_STRING_ENCODING_IS_ASCII || __PYX_DEFAULT_STRING_ENCODING_IS_DEFAULT + if ( +#if PY_MAJOR_VERSION < 3 && __PYX_DEFAULT_STRING_ENCODING_IS_ASCII + __Pyx_sys_getdefaultencoding_not_ascii && +#endif + PyUnicode_Check(o)) { + return __Pyx_PyUnicode_AsStringAndSize(o, length); + } else +#endif +#if (!CYTHON_COMPILING_IN_PYPY) || (defined(PyByteArray_AS_STRING) && defined(PyByteArray_GET_SIZE)) + if (PyByteArray_Check(o)) { + *length = PyByteArray_GET_SIZE(o); + return PyByteArray_AS_STRING(o); + } else +#endif + { + char* result; + int r = PyBytes_AsStringAndSize(o, &result, length); + if (unlikely(r < 0)) { + return NULL; + } else { + return result; + } + } +} +static CYTHON_INLINE int __Pyx_PyObject_IsTrue(PyObject* x) { + int is_true = x == Py_True; + if (is_true | (x == Py_False) | (x == Py_None)) return is_true; + else return PyObject_IsTrue(x); +} +static CYTHON_INLINE int __Pyx_PyObject_IsTrueAndDecref(PyObject* x) { + int retval; + if (unlikely(!x)) return -1; + retval = __Pyx_PyObject_IsTrue(x); + Py_DECREF(x); + return retval; +} +static PyObject* __Pyx_PyNumber_IntOrLongWrongResultType(PyObject* result, const char* type_name) { +#if PY_MAJOR_VERSION >= 3 + if (PyLong_Check(result)) { + if (PyErr_WarnFormat(PyExc_DeprecationWarning, 1, + "__int__ returned non-int (type %.200s). " + "The ability to return an instance of a strict subclass of int " + "is deprecated, and may be removed in a future version of Python.", + Py_TYPE(result)->tp_name)) { + Py_DECREF(result); + return NULL; + } + return result; + } +#endif + PyErr_Format(PyExc_TypeError, + "__%.4s__ returned non-%.4s (type %.200s)", + type_name, type_name, Py_TYPE(result)->tp_name); + Py_DECREF(result); + return NULL; +} +static CYTHON_INLINE PyObject* __Pyx_PyNumber_IntOrLong(PyObject* x) { +#if CYTHON_USE_TYPE_SLOTS + PyNumberMethods *m; +#endif + const char *name = NULL; + PyObject *res = NULL; +#if PY_MAJOR_VERSION < 3 + if (likely(PyInt_Check(x) || PyLong_Check(x))) +#else + if (likely(PyLong_Check(x))) +#endif + return __Pyx_NewRef(x); +#if CYTHON_USE_TYPE_SLOTS + m = Py_TYPE(x)->tp_as_number; + #if PY_MAJOR_VERSION < 3 + if (m && m->nb_int) { + name = "int"; + res = m->nb_int(x); + } + else if (m && m->nb_long) { + name = "long"; + res = m->nb_long(x); + } + #else + if (likely(m && m->nb_int)) { + name = "int"; + res = m->nb_int(x); + } + #endif +#else + if (!PyBytes_CheckExact(x) && !PyUnicode_CheckExact(x)) { + res = PyNumber_Int(x); + } +#endif + if (likely(res)) { +#if PY_MAJOR_VERSION < 3 + if (unlikely(!PyInt_Check(res) && !PyLong_Check(res))) { +#else + if (unlikely(!PyLong_CheckExact(res))) { +#endif + return __Pyx_PyNumber_IntOrLongWrongResultType(res, name); + } + } + else if (!PyErr_Occurred()) { + PyErr_SetString(PyExc_TypeError, + "an integer is required"); + } + return res; +} +static CYTHON_INLINE Py_ssize_t __Pyx_PyIndex_AsSsize_t(PyObject* b) { + Py_ssize_t ival; + PyObject *x; +#if PY_MAJOR_VERSION < 3 + if (likely(PyInt_CheckExact(b))) { + if (sizeof(Py_ssize_t) >= sizeof(long)) + return PyInt_AS_LONG(b); + else + return PyInt_AsSsize_t(b); + } +#endif + if (likely(PyLong_CheckExact(b))) { + #if CYTHON_USE_PYLONG_INTERNALS + const digit* digits = ((PyLongObject*)b)->ob_digit; + const Py_ssize_t size = Py_SIZE(b); + if (likely(__Pyx_sst_abs(size) <= 1)) { + ival = likely(size) ? digits[0] : 0; + if (size == -1) ival = -ival; + return ival; + } else { + switch (size) { + case 2: + if (8 * sizeof(Py_ssize_t) > 2 * PyLong_SHIFT) { + return (Py_ssize_t) (((((size_t)digits[1]) << PyLong_SHIFT) | (size_t)digits[0])); + } + break; + case -2: + if (8 * sizeof(Py_ssize_t) > 2 * PyLong_SHIFT) { + return -(Py_ssize_t) (((((size_t)digits[1]) << PyLong_SHIFT) | (size_t)digits[0])); + } + break; + case 3: + if (8 * sizeof(Py_ssize_t) > 3 * PyLong_SHIFT) { + return (Py_ssize_t) (((((((size_t)digits[2]) << PyLong_SHIFT) | (size_t)digits[1]) << PyLong_SHIFT) | (size_t)digits[0])); + } + break; + case -3: + if (8 * sizeof(Py_ssize_t) > 3 * PyLong_SHIFT) { + return -(Py_ssize_t) (((((((size_t)digits[2]) << PyLong_SHIFT) | (size_t)digits[1]) << PyLong_SHIFT) | (size_t)digits[0])); + } + break; + case 4: + if (8 * sizeof(Py_ssize_t) > 4 * PyLong_SHIFT) { + return (Py_ssize_t) (((((((((size_t)digits[3]) << PyLong_SHIFT) | (size_t)digits[2]) << PyLong_SHIFT) | (size_t)digits[1]) << PyLong_SHIFT) | (size_t)digits[0])); + } + break; + case -4: + if (8 * sizeof(Py_ssize_t) > 4 * PyLong_SHIFT) { + return -(Py_ssize_t) (((((((((size_t)digits[3]) << PyLong_SHIFT) | (size_t)digits[2]) << PyLong_SHIFT) | (size_t)digits[1]) << PyLong_SHIFT) | (size_t)digits[0])); + } + break; + } + } + #endif + return PyLong_AsSsize_t(b); + } + x = PyNumber_Index(b); + if (!x) return -1; + ival = PyInt_AsSsize_t(x); + Py_DECREF(x); + return ival; +} +static CYTHON_INLINE Py_hash_t __Pyx_PyIndex_AsHash_t(PyObject* o) { + if (sizeof(Py_hash_t) == sizeof(Py_ssize_t)) { + return (Py_hash_t) __Pyx_PyIndex_AsSsize_t(o); +#if PY_MAJOR_VERSION < 3 + } else if (likely(PyInt_CheckExact(o))) { + return PyInt_AS_LONG(o); +#endif + } else { + Py_ssize_t ival; + PyObject *x; + x = PyNumber_Index(o); + if (!x) return -1; + ival = PyInt_AsLong(x); + Py_DECREF(x); + return ival; + } +} +static CYTHON_INLINE PyObject * __Pyx_PyBool_FromLong(long b) { + return b ? __Pyx_NewRef(Py_True) : __Pyx_NewRef(Py_False); +} +static CYTHON_INLINE PyObject * __Pyx_PyInt_FromSize_t(size_t ival) { + return PyInt_FromSize_t(ival); +} + + +#endif /* Py_PYTHON_H */ diff --git a/models/cv/detection/retinaface/ixrt/box_overlaps.pyx b/models/cv/detection/retinaface/ixrt/box_overlaps.pyx new file mode 100644 index 0000000000000000000000000000000000000000..c5ae69d730f657350320bcf782df458c8c7e1637 --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/box_overlaps.pyx @@ -0,0 +1,57 @@ +# -------------------------------------------------------- +# Fast R-CNN +# Copyright (c) 2015 Microsoft +# Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# Licensed under The MIT License [see LICENSE for details] +# Written by Sergey Karayev +# -------------------------------------------------------- + +cimport cython +import numpy as np +cimport numpy as np + +DTYPE = float +ctypedef np.float_t DTYPE_t + +def bbox_overlaps( + np.ndarray[DTYPE_t, ndim=2] boxes, + np.ndarray[DTYPE_t, ndim=2] query_boxes): + """ + Parameters + ---------- + boxes: (N, 4) ndarray of float + query_boxes: (K, 4) ndarray of float + Returns + ------- + overlaps: (N, K) ndarray of overlap between boxes and query_boxes + """ + cdef unsigned int N = boxes.shape[0] + cdef unsigned int K = query_boxes.shape[0] + cdef np.ndarray[DTYPE_t, ndim=2] overlaps = np.zeros((N, K), dtype=DTYPE) + cdef DTYPE_t iw, ih, box_area + cdef DTYPE_t ua + cdef unsigned int k, n + for k in range(K): + box_area = ( + (query_boxes[k, 2] - query_boxes[k, 0] + 1) * + (query_boxes[k, 3] - query_boxes[k, 1] + 1) + ) + for n in range(N): + iw = ( + min(boxes[n, 2], query_boxes[k, 2]) - + max(boxes[n, 0], query_boxes[k, 0]) + 1 + ) + if iw > 0: + ih = ( + min(boxes[n, 3], query_boxes[k, 3]) - + max(boxes[n, 1], query_boxes[k, 1]) + 1 + ) + if ih > 0: + ua = float( + (boxes[n, 2] - boxes[n, 0] + 1) * + (boxes[n, 3] - boxes[n, 1] + 1) + + box_area - iw * ih + ) + overlaps[n, k] = iw * ih / ua + return overlaps \ No newline at end of file diff --git a/models/cv/detection/retinaface/ixrt/build_engine.py b/models/cv/detection/retinaface/ixrt/build_engine.py new file mode 100644 index 0000000000000000000000000000000000000000..e0a3bb4bcd143c68406d915245eefb7060010127 --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/build_engine.py @@ -0,0 +1,53 @@ +# Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import cv2 +import argparse +import numpy as np + +import torch +import tensorrt + +def main(config): + IXRT_LOGGER = tensorrt.Logger(tensorrt.Logger.WARNING) + builder = tensorrt.Builder(IXRT_LOGGER) + EXPLICIT_BATCH = 1 << (int)(tensorrt.NetworkDefinitionCreationFlag.EXPLICIT_BATCH) + network = builder.create_network(EXPLICIT_BATCH) + build_config = builder.create_builder_config() + parser = tensorrt.OnnxParser(network, IXRT_LOGGER) + parser.parse_from_file(config.model) + + precision = tensorrt.BuilderFlag.INT8 if config.precision == "int8" else tensorrt.BuilderFlag.FP16 + # print("precision : ", precision) + build_config.set_flag(precision) + + plan = builder.build_serialized_network(network, build_config) + engine_file_path = config.engine + with open(engine_file_path, "wb") as f: + f.write(plan) + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--model", type=str) + parser.add_argument("--precision", type=str, choices=["float16", "int8", "float32"], default="int8", + help="The precision of datatype") + parser.add_argument("--engine", type=str, default=None) + args = parser.parse_args() + return args + +if __name__ == "__main__": + args = parse_args() + main(args) \ No newline at end of file diff --git a/models/cv/detection/retinaface/ixrt/common.py b/models/cv/detection/retinaface/ixrt/common.py new file mode 100644 index 0000000000000000000000000000000000000000..1d7f96e8a5df6b4db9a882d8d29a7fc4a2630702 --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/common.py @@ -0,0 +1,69 @@ +# Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import cv2 +import glob +import torch +import tensorrt +import numpy as np +from cuda import cuda, cudart + +def create_engine_context(engine_path, logger): + with open(engine_path, "rb") as f: + runtime = tensorrt.Runtime(logger) + assert runtime + engine = runtime.deserialize_cuda_engine(f.read()) + assert engine + context = engine.create_execution_context() + assert context + + return engine, context + +def get_io_bindings(engine): + # Setup I/O bindings + inputs = [] + outputs = [] + allocations = [] + + for i in range(engine.num_bindings): + is_input = False + if engine.binding_is_input(i): + is_input = True + name = engine.get_binding_name(i) + dtype = engine.get_binding_dtype(i) + shape = engine.get_binding_shape(i) + if is_input: + batch_size = shape[0] + size = np.dtype(tensorrt.nptype(dtype)).itemsize + for s in shape: + size *= s + err, allocation = cudart.cudaMalloc(size) + assert err == cudart.cudaError_t.cudaSuccess + binding = { + "index": i, + "name": name, + "dtype": np.dtype(tensorrt.nptype(dtype)), + "shape": list(shape), + "allocation": allocation, + "nbytes": size, + } + print(f"binding {i}, name : {name} dtype : {np.dtype(tensorrt.nptype(dtype))} shape : {list(shape)}") + allocations.append(allocation) + if engine.binding_is_input(i): + inputs.append(binding) + else: + outputs.append(binding) + return inputs, outputs, allocations diff --git a/models/cv/detection/retinaface/ixrt/config/RETINAFACE_CONFIG b/models/cv/detection/retinaface/ixrt/config/RETINAFACE_CONFIG new file mode 100644 index 0000000000000000000000000000000000000000..b10fe9754cae67ce592180f407517d5b489bc50e --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/config/RETINAFACE_CONFIG @@ -0,0 +1,34 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +# IMGSIZE : 模型输入hw大小 +# MODEL_NAME : 生成onnx/engine的basename +# ORIGINE_MODEL : 原始onnx文件名称 +IMGSIZE=320 +MODEL_NAME=RetinaFace +ORIGINE_MODEL=mnetv1_retinaface.onnx + +# QUANT CONFIG (仅PRECISION为int8时生效) + # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] + # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) + # QUANT_STEP : 量化步数 + # QUANT_SEED : 随机种子 保证量化结果可复现 + # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 +QUANT_OBSERVER=hist_percentile +QUANT_BATCHSIZE=32 +QUANT_STEP=32 +QUANT_SEED=42 +DISABLE_QUANT_LIST= +QUANT_EXIST_ONNX= diff --git a/models/cv/detection/retinaface/ixrt/evaluation.py b/models/cv/detection/retinaface/ixrt/evaluation.py new file mode 100644 index 0000000000000000000000000000000000000000..d3c6495aaf782725761a52c9d2442564b482155d --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/evaluation.py @@ -0,0 +1,293 @@ +# Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import tqdm +import pickle +import argparse +import numpy as np +from scipy.io import loadmat +from bbox import bbox_overlaps + + +def get_gt_boxes(gt_dir): + """ gt dir: (wider_face_val.mat, wider_easy_val.mat, wider_medium_val.mat, wider_hard_val.mat)""" + + gt_mat = loadmat(os.path.join(gt_dir, 'wider_face_val.mat')) + hard_mat = loadmat(os.path.join(gt_dir, 'wider_hard_val.mat')) + medium_mat = loadmat(os.path.join(gt_dir, 'wider_medium_val.mat')) + easy_mat = loadmat(os.path.join(gt_dir, 'wider_easy_val.mat')) + + facebox_list = gt_mat['face_bbx_list'] + event_list = gt_mat['event_list'] + file_list = gt_mat['file_list'] + + hard_gt_list = hard_mat['gt_list'] + medium_gt_list = medium_mat['gt_list'] + easy_gt_list = easy_mat['gt_list'] + + return facebox_list, event_list, file_list, hard_gt_list, medium_gt_list, easy_gt_list + + +def get_gt_boxes_from_txt(gt_path, cache_dir): + + cache_file = os.path.join(cache_dir, 'gt_cache.pkl') + if os.path.exists(cache_file): + f = open(cache_file, 'rb') + boxes = pickle.load(f) + f.close() + return boxes + + f = open(gt_path, 'r') + state = 0 + lines = f.readlines() + lines = list(map(lambda x: x.rstrip('\r\n'), lines)) + boxes = {} + print(len(lines)) + f.close() + current_boxes = [] + current_name = None + for line in lines: + if state == 0 and '--' in line: + state = 1 + current_name = line + continue + if state == 1: + state = 2 + continue + + if state == 2 and '--' in line: + state = 1 + boxes[current_name] = np.array(current_boxes).astype('float32') + current_name = line + current_boxes = [] + continue + + if state == 2: + box = [float(x) for x in line.split(' ')[:4]] + current_boxes.append(box) + continue + + f = open(cache_file, 'wb') + pickle.dump(boxes, f) + f.close() + return boxes + + +def read_pred_file(filepath): + + with open(filepath, 'r') as f: + lines = f.readlines() + img_file = lines[0].rstrip('\n\r') + lines = lines[2:] + + # b = lines[0].rstrip('\r\n').split(' ')[:-1] + # c = float(b) + # a = map(lambda x: [[float(a[0]), float(a[1]), float(a[2]), float(a[3]), float(a[4])] for a in x.rstrip('\r\n').split(' ')], lines) + boxes = [] + for line in lines: + line = line.rstrip('\r\n').split(' ') + if line[0] is '': + continue + # a = float(line[4]) + boxes.append([float(line[0]), float(line[1]), float(line[2]), float(line[3]), float(line[4])]) + boxes = np.array(boxes) + # boxes = np.array(list(map(lambda x: [float(a) for a in x.rstrip('\r\n').split(' ')], lines))).astype('float') + return img_file.split('/')[-1], boxes + + +def get_preds(pred_dir): + events = os.listdir(pred_dir) + boxes = dict() + pbar = tqdm.tqdm(events) + + for event in pbar: + pbar.set_description('Reading Predictions ') + event_dir = os.path.join(pred_dir, event) + event_images = os.listdir(event_dir) + current_event = dict() + for imgtxt in event_images: + imgname, _boxes = read_pred_file(os.path.join(event_dir, imgtxt)) + current_event[imgname.rstrip('.jpg')] = _boxes + boxes[event] = current_event + return boxes + + +def norm_score(pred): + """ norm score + pred {key: [[x1,y1,x2,y2,s]]} + """ + + max_score = 0 + min_score = 1 + + for _, k in pred.items(): + for _, v in k.items(): + if len(v) == 0: + continue + _min = np.min(v[:, -1]) + _max = np.max(v[:, -1]) + max_score = max(_max, max_score) + min_score = min(_min, min_score) + + diff = max_score - min_score + for _, k in pred.items(): + for _, v in k.items(): + if len(v) == 0: + continue + v[:, -1] = (v[:, -1] - min_score)/diff + + +def image_eval(pred, gt, ignore, iou_thresh): + """ single image evaluation + pred: Nx5 + gt: Nx4 + ignore: + """ + + _pred = pred.copy() + _gt = gt.copy() + pred_recall = np.zeros(_pred.shape[0]) + recall_list = np.zeros(_gt.shape[0]) + proposal_list = np.ones(_pred.shape[0]) + + _pred[:, 2] = _pred[:, 2] + _pred[:, 0] + _pred[:, 3] = _pred[:, 3] + _pred[:, 1] + _gt[:, 2] = _gt[:, 2] + _gt[:, 0] + _gt[:, 3] = _gt[:, 3] + _gt[:, 1] + + overlaps = bbox_overlaps(_pred[:, :4], _gt) + + for h in range(_pred.shape[0]): + + gt_overlap = overlaps[h] + max_overlap, max_idx = gt_overlap.max(), gt_overlap.argmax() + if max_overlap >= iou_thresh: + if ignore[max_idx] == 0: + recall_list[max_idx] = -1 + proposal_list[h] = -1 + elif recall_list[max_idx] == 0: + recall_list[max_idx] = 1 + + r_keep_index = np.where(recall_list == 1)[0] + pred_recall[h] = len(r_keep_index) + return pred_recall, proposal_list + + +def img_pr_info(thresh_num, pred_info, proposal_list, pred_recall): + pr_info = np.zeros((thresh_num, 2)).astype('float') + for t in range(thresh_num): + + thresh = 1 - (t+1)/thresh_num + r_index = np.where(pred_info[:, 4] >= thresh)[0] + if len(r_index) == 0: + pr_info[t, 0] = 0 + pr_info[t, 1] = 0 + else: + r_index = r_index[-1] + p_index = np.where(proposal_list[:r_index+1] == 1)[0] + pr_info[t, 0] = len(p_index) + pr_info[t, 1] = pred_recall[r_index] + return pr_info + + +def dataset_pr_info(thresh_num, pr_curve, count_face): + _pr_curve = np.zeros((thresh_num, 2)) + for i in range(thresh_num): + _pr_curve[i, 0] = pr_curve[i, 1] / pr_curve[i, 0] + _pr_curve[i, 1] = pr_curve[i, 1] / count_face + return _pr_curve + + +def voc_ap(rec, prec): + + # correct AP calculation + # first append sentinel values at the end + mrec = np.concatenate(([0.], rec, [1.])) + mpre = np.concatenate(([0.], prec, [0.])) + + # compute the precision envelope + for i in range(mpre.size - 1, 0, -1): + mpre[i - 1] = np.maximum(mpre[i - 1], mpre[i]) + + # to calculate area under PR curve, look for points + # where X axis (recall) changes value + i = np.where(mrec[1:] != mrec[:-1])[0] + + # and sum (\Delta recall) * prec + ap = np.sum((mrec[i + 1] - mrec[i]) * mpre[i + 1]) + return ap + + +def evaluation(pred, gt_path, iou_thresh=0.5): + pred = get_preds(pred) + norm_score(pred) + facebox_list, event_list, file_list, hard_gt_list, medium_gt_list, easy_gt_list = get_gt_boxes(gt_path) + event_num = len(event_list) + thresh_num = 1000 + settings = ['easy', 'medium', 'hard'] + setting_gts = [easy_gt_list, medium_gt_list, hard_gt_list] + aps = [] + for setting_id in range(3): + # different setting + gt_list = setting_gts[setting_id] + count_face = 0 + pr_curve = np.zeros((thresh_num, 2)).astype('float') + # [hard, medium, easy] + pbar = tqdm.tqdm(range(event_num)) + for i in pbar: + pbar.set_description('Processing {}'.format(settings[setting_id])) + event_name = str(event_list[i][0][0]) + img_list = file_list[i][0] + pred_list = pred[event_name] + sub_gt_list = gt_list[i][0] + # img_pr_info_list = np.zeros((len(img_list), thresh_num, 2)) + gt_bbx_list = facebox_list[i][0] + + for j in range(len(img_list)): + if str(img_list[j][0][0]) not in pred_list: + continue + + pred_info = pred_list[str(img_list[j][0][0])] + + gt_boxes = gt_bbx_list[j][0].astype('float') + keep_index = sub_gt_list[j][0] + count_face += len(keep_index) + + if len(gt_boxes) == 0 or len(pred_info) == 0: + continue + ignore = np.zeros(gt_boxes.shape[0]) + if len(keep_index) != 0: + ignore[keep_index-1] = 1 + pred_recall, proposal_list = image_eval(pred_info, gt_boxes, ignore, iou_thresh) + + _img_pr_info = img_pr_info(thresh_num, pred_info, proposal_list, pred_recall) + + pr_curve += _img_pr_info + pr_curve = dataset_pr_info(thresh_num, pr_curve, count_face) + + propose = pr_curve[:, 0] + recall = pr_curve[:, 1] + + ap = voc_ap(recall, propose) + aps.append(ap) + + print("==================== Results ====================") + print("Easy Val AP: {}".format(aps[0])) + print("Medium Val AP: {}".format(aps[1])) + print("Hard Val AP: {}".format(aps[2])) + print("=================================================") + + return aps[0] \ No newline at end of file diff --git a/models/cv/detection/retinaface/ixrt/inference.py b/models/cv/detection/retinaface/ixrt/inference.py new file mode 100644 index 0000000000000000000000000000000000000000..7c93103cf154e9fb2f46af9d685084ed357c89c6 --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/inference.py @@ -0,0 +1,289 @@ +#!/usr/bin/env python +# -*- coding: utf-8 -*- +# Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import argparse +import json +import os +import re +import time +from tqdm import tqdm + +import cv2 +import numpy as np +from cuda import cuda, cudart +import torch +import onnx +import onnxruntime +import torchvision.datasets +from torch.utils.data import DataLoader +import torch.nn.functional as F +import tensorrt + +from wider_face_dataset import WiderFaceDetection, detection_collate +from models import * +from evaluation import evaluation +from common import create_engine_context, get_io_bindings + +def post_process(args, prior_data, locs, confs, landms, resizes, img_files, net_h, net_w): + import numpy as np + from utils.box_utils import decode, decode_landm + from utils.nms.py_cpu_nms import py_cpu_nms + + variances = 0.1, 0.2 + nms_threshold = 0.4 + confidence_threshold = 0.02 + + scale = torch.Tensor([net_w, net_h, net_w, net_h]) + scale1 = torch.Tensor([net_w, net_h, net_w, net_h, + net_w, net_h, net_w, net_h, + net_w, net_h]) + + for i, (loc, conf, landm, resize, img_name) in enumerate(zip(locs, confs, landms, resizes, img_files)): + boxes = decode(loc.squeeze(0).data, prior_data, variances) + boxes = boxes * scale / resize + boxes = boxes.cpu().numpy() + + scores = conf.squeeze(0).data.cpu().numpy()[:, 1] + landm = decode_landm(landm.squeeze(0).data, prior_data, variances) + + landm = landm * scale1 / resize + landm = landm.cpu().numpy() + + # ignore low scores + inds = np.where(scores > confidence_threshold)[0] + boxes = boxes[inds] + landm = landm[inds] + scores = scores[inds] + + order = scores.argsort()[::-1] + boxes = boxes[order] + landm = landm[order] + scores = scores[order] + + # do NMS + dets = np.hstack((boxes, scores[:, np.newaxis])).astype(np.float32, copy=False) + keep = py_cpu_nms(dets, nms_threshold) + dets = dets[keep, :] + landm = landm[keep] + + dets = np.concatenate((dets, landm), axis=1) + + # -------------------------------------------------------------------- + save_name = os.path.join(args.save_folder, f"{img_name[:-4]}.txt") + dirname = os.path.dirname(save_name) + if not os.path.isdir(dirname): + os.makedirs(dirname) + + with open(save_name, "w") as fd: + bboxs = dets + file_name = os.path.basename(save_name)[:-4] + "\n" + bboxs_num = str(len(bboxs)) + "\n" + fd.write(file_name) + fd.write(bboxs_num) + for box in bboxs: + x = int(box[0]) + y = int(box[1]) + w = int(box[2]) - int(box[0]) + h = int(box[3]) - int(box[1]) + confidence = str(box[4]) + line = str(x) + " " + str(y) + " " + str(w) + " " + str(h) + " " + confidence + " \n" + fd.write(line) + + +def main(config): + dataset = WiderFaceDetection(config.datasets_dir) + + dataloader = DataLoader( + dataset, + shuffle=False, + batch_size=config.bsz, + drop_last=True, + num_workers=4, + collate_fn=detection_collate + ) + + host_mem = tensorrt.IHostMemory + logger = tensorrt.Logger(tensorrt.Logger.ERROR) + + # Load Engine && I/O bindings + engine, context = create_engine_context(config.engine_file, logger) + inputs, outputs, allocations = get_io_bindings(engine) + + # Warm up + if config.warm_up > 0: + print("\nWarm Start.") + for i in range(config.warm_up): + context.execute_v2(allocations) + print("Warm Done.") + + # Inference + if config.test_mode == "FPS": + torch.cuda.synchronize() + start_time = time.time() + + for i in range(config.loop_count): + context.execute_v2(allocations) + + torch.cuda.synchronize() + end_time = time.time() + forward_time = end_time - start_time + + fps = config.loop_count * config.bsz / forward_time + + print("FPS : ", fps) + print(f"Performance Check : Test {fps} >= target {config.fps_target}") + if fps >= config.fps_target: + print("pass!") + exit() + else: + print("failed!") + exit(1) + + elif config.test_mode == "ACC": + + ## Prepare the output data + bbox_out0 = np.zeros(outputs[0]["shape"], outputs[0]["dtype"]) + + bbox_out1 = np.zeros(outputs[3]["shape"], outputs[3]["dtype"]) + + bbox_out2 = np.zeros(outputs[6]["shape"], outputs[6]["dtype"]) + + cls_out0 = np.zeros(outputs[1]["shape"], outputs[1]["dtype"]) + + cls_out1 = np.zeros(outputs[4]["shape"], outputs[4]["dtype"]) + + cls_out2 = np.zeros(outputs[7]["shape"], outputs[7]["dtype"]) + + ldm_out0 = np.zeros(outputs[2]["shape"], outputs[2]["dtype"]) + + ldm_out1 = np.zeros(outputs[5]["shape"], outputs[5]["dtype"]) + + ldm_out2 = np.zeros(outputs[8]["shape"], outputs[8]["dtype"]) + + + priorbox = PriorBox(image_size=(320, 320)) + priors = priorbox.forward() + prior_data = priors.data + + for i, data in enumerate(tqdm(dataloader)): + + #print(i) + imgs, rs, img_files = data + batch_data = imgs.numpy().astype(inputs[0]["dtype"]) + batch_data = np.ascontiguousarray(batch_data) + + err, = cuda.cuMemcpyHtoD(inputs[0]["allocation"], batch_data, batch_data.nbytes) + assert(err == cuda.CUresult.CUDA_SUCCESS) + context.execute_v2(allocations) + + err, = cuda.cuMemcpyDtoH(bbox_out0, outputs[0]["allocation"], outputs[0]["nbytes"]) + assert(err == cuda.CUresult.CUDA_SUCCESS) + bbox_out0_t = torch.from_numpy(bbox_out0.transpose([0,2,3,1])).reshape(bbox_out0.shape[0], -1, 4) + + err, = cuda.cuMemcpyDtoH(bbox_out1, outputs[3]["allocation"], outputs[3]["nbytes"]) + assert(err == cuda.CUresult.CUDA_SUCCESS) + bbox_out1_t = torch.from_numpy(bbox_out1.transpose([0,2,3,1])).reshape(bbox_out1.shape[0], -1, 4) + + err, = cuda.cuMemcpyDtoH(bbox_out2, outputs[6]["allocation"], outputs[6]["nbytes"]) + assert(err == cuda.CUresult.CUDA_SUCCESS) + bbox_out2_t = torch.from_numpy(bbox_out2.transpose([0,2,3,1])).reshape(bbox_out2.shape[0], -1, 4) + + err, = cuda.cuMemcpyDtoH(cls_out0, outputs[1]["allocation"], outputs[1]["nbytes"]) + assert(err == cuda.CUresult.CUDA_SUCCESS) + cls_out0_t = torch.from_numpy(cls_out0.transpose([0,2,3,1])).reshape(cls_out0.shape[0], -1, 2) + + err, = cuda.cuMemcpyDtoH(cls_out1, outputs[4]["allocation"], outputs[4]["nbytes"]) + assert(err == cuda.CUresult.CUDA_SUCCESS) + cls_out1_t = torch.from_numpy(cls_out1.transpose([0,2,3,1])).reshape(cls_out1.shape[0], -1, 2) + + err, = cuda.cuMemcpyDtoH(cls_out2, outputs[7]["allocation"], outputs[7]["nbytes"]) + assert(err == cuda.CUresult.CUDA_SUCCESS) + cls_out2_t = torch.from_numpy(cls_out2.transpose([0,2,3,1])).reshape(cls_out2.shape[0], -1, 2) + + err, = cuda.cuMemcpyDtoH(ldm_out0, outputs[2]["allocation"], outputs[2]["nbytes"]) + assert(err == cuda.CUresult.CUDA_SUCCESS) + ldm_out0_t = torch.from_numpy(ldm_out0.transpose([0,2,3,1])).reshape(ldm_out0.shape[0], -1, 10) + + err, = cuda.cuMemcpyDtoH(ldm_out1, outputs[5]["allocation"], outputs[5]["nbytes"]) + assert(err == cuda.CUresult.CUDA_SUCCESS) + ldm_out1_t = torch.from_numpy(ldm_out1.transpose([0,2,3,1])).reshape(ldm_out1.shape[0], -1, 10) + + err, = cuda.cuMemcpyDtoH(ldm_out2, outputs[8]["allocation"], outputs[8]["nbytes"]) + assert(err == cuda.CUresult.CUDA_SUCCESS) + ldm_out2_t = torch.from_numpy(ldm_out2.transpose([0,2,3,1])).reshape(ldm_out2.shape[0], -1, 10) + + + + bbox_regressions = torch.cat([bbox_out0_t, bbox_out1_t, bbox_out2_t], dim=1) + classifications = torch.cat([cls_out0_t, cls_out1_t, cls_out2_t], dim=1) + ldm_regressions = torch.cat([ldm_out0_t, ldm_out1_t, ldm_out2_t], dim=1) + classifications = F.softmax(classifications, dim=-1) + + net_h, net_w = batch_data.shape[2], batch_data.shape[3] + + post_process(config, prior_data, bbox_regressions, classifications, ldm_regressions, rs, img_files, net_h, net_w) + + easy_AP = evaluation(config.pred_path, config.gt) + + if easy_AP >= config.acc_target: + print("pass!") + exit() + else: + print("failed!") + exit(1) + +def parse_config(): + parser = argparse.ArgumentParser() + parser.add_argument("--test_mode", type=str, default="FPS", help="FPS MAP") + parser.add_argument( + "--engine_file", + type=str, + help="engine file path" + ) + parser.add_argument( + "--datasets_dir", + type=str, + default="", + help="ImageNet dir", + ) + parser.add_argument("--warm_up", type=int, default=-1, help="warm_up times") + parser.add_argument("--bsz", type=int, default=32, help="test batch size") + parser.add_argument( + "--imgsz", + "--img", + "--img-size", + type=int, + default=224, + help="inference size h,w", + ) + parser.add_argument("--save_folder", type=str, default=None) + parser.add_argument("--pred_path", type=str, default=None) + parser.add_argument('-g', '--gt', default=None) + parser.add_argument("--use_async", action="store_true") + parser.add_argument( + "--device", type=int, default=0, help="cuda device, i.e. 0 or 0,1,2,3,4" + ) + parser.add_argument("--fps_target", type=float, default=-1.0) + parser.add_argument("--acc_target", type=float, default=-1.0) + parser.add_argument("--loop_count", type=int, default=-1) + + config = parser.parse_args() + return config + +if __name__ == "__main__": + config = parse_config() + main(config) diff --git a/models/cv/detection/retinaface/ixrt/models/__init__.py b/models/cv/detection/retinaface/ixrt/models/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..8e7f9e30838fd1e93b6b58a96877958aaf31f54b --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/models/__init__.py @@ -0,0 +1,26 @@ +# Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +from .config import cfg_mnet +from .prior_box import PriorBox +from .net import MobileNetV1 +from .retinaface import mnetv1_retinaface + +__all__ = [ + "cfg_mnet", + "PriorBox", + "MobileNetV1", + "mnetv1_retinaface" +] \ No newline at end of file diff --git a/models/cv/detection/retinaface/ixrt/models/config.py b/models/cv/detection/retinaface/ixrt/models/config.py new file mode 100644 index 0000000000000000000000000000000000000000..ea105db62daa9b3a86b8598125a9bdc79b086379 --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/models/config.py @@ -0,0 +1,21 @@ +# Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +cfg_mnet = { + 'min_sizes': [[10, 20], [32, 64], [128, 256]], + 'steps': [8, 16, 32], + 'clip': False, + 'variance': [0.1, 0.2] +} \ No newline at end of file diff --git a/models/cv/detection/retinaface/ixrt/models/net.py b/models/cv/detection/retinaface/ixrt/models/net.py new file mode 100644 index 0000000000000000000000000000000000000000..30a095283adb9da64704fe3fe88c6abdd3ab5cb9 --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/models/net.py @@ -0,0 +1,140 @@ +# Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +from typing import Tuple +import torch +import torch.nn as nn +import torch.nn.functional as F + +def conv_bn(inp, oup, stride = 1): + return nn.Sequential( + nn.Conv2d(inp, oup, 3, stride, 1, bias=False), + nn.BatchNorm2d(oup), + nn.ReLU(inplace=True) + ) + +def conv_bn_no_relu(inp, oup, stride): + return nn.Sequential( + nn.Conv2d(inp, oup, 3, stride, 1, bias=False), + nn.BatchNorm2d(oup), + ) + +def conv_bn1X1(inp, oup, stride): + return nn.Sequential( + nn.Conv2d(inp, oup, 1, stride, padding=0, bias=False), + nn.BatchNorm2d(oup), + nn.ReLU(inplace=True) + ) + +def conv_dw(inp, oup, stride): + return nn.Sequential( + nn.Conv2d(inp, inp, 3, stride, 1, groups=inp, bias=False), + nn.BatchNorm2d(inp), + nn.ReLU(inplace=True), + + nn.Conv2d(inp, oup, 1, 1, 0, bias=False), + nn.BatchNorm2d(oup), + nn.ReLU(inplace=True) + ) + +class SSH(nn.Module): + def __init__(self, in_channel, out_channel): + super(SSH, self).__init__() + assert out_channel % 4 == 0 + leaky = 0 + if (out_channel <= 64): + leaky = 0.1 + self.conv3X3 = conv_bn_no_relu(in_channel, out_channel//2, stride=1) + + self.conv5X5_1 = conv_bn(in_channel, out_channel//4, stride=1) + self.conv5X5_2 = conv_bn_no_relu(out_channel//4, out_channel//4, stride=1) + + self.conv7X7_2 = conv_bn(out_channel//4, out_channel//4, stride=1) + self.conv7x7_3 = conv_bn_no_relu(out_channel//4, out_channel//4, stride=1) + + def forward(self, input): + conv3X3 = self.conv3X3(input) + + conv5X5_1 = self.conv5X5_1(input) + conv5X5 = self.conv5X5_2(conv5X5_1) + + conv7X7_2 = self.conv7X7_2(conv5X5_1) + conv7X7 = self.conv7x7_3(conv7X7_2) + + out = torch.cat([conv3X3, conv5X5, conv7X7], dim=1) + out = F.relu(out) + return out + +class FPN(nn.Module): + def __init__(self,in_channels_list,out_channels): + super(FPN,self).__init__() + leaky = 0 + if (out_channels <= 64): + leaky = 0.1 + self.output1 = conv_bn1X1(in_channels_list[0], out_channels, stride = 1) + self.output2 = conv_bn1X1(in_channels_list[1], out_channels, stride = 1) + self.output3 = conv_bn1X1(in_channels_list[2], out_channels, stride = 1) + + self.merge1 = conv_bn(out_channels, out_channels) + self.merge2 = conv_bn(out_channels, out_channels) + + def forward(self, input: Tuple[torch.Tensor, torch.Tensor, torch.Tensor]): + output1 = self.output1(input[0]) + output2 = self.output2(input[1]) + output3 = self.output3(input[2]) + + up3 = F.interpolate(output3, size=[output2.size(2), output2.size(3)], mode="nearest") + output2 = output2 + up3 + output2 = self.merge2(output2) + + up2 = F.interpolate(output2, size=[output1.size(2), output1.size(3)], mode="nearest") + output1 = output1 + up2 + output1 = self.merge1(output1) + + out = [output1, output2, output3] + return out + + + +class MobileNetV1(nn.Module): + def __init__(self): + super(MobileNetV1, self).__init__() + self.stage1 = nn.Sequential( + conv_bn(3, 8, 2), # 3 + conv_dw(8, 16, 1), # 7 + conv_dw(16, 32, 2), # 11 + conv_dw(32, 32, 1), # 19 + conv_dw(32, 64, 2), # 27 + conv_dw(64, 64, 1), # 43 + ) + self.stage2 = nn.Sequential( + conv_dw(64, 128, 2), # 43 + 16 = 59 + conv_dw(128, 128, 1), # 59 + 32 = 91 + conv_dw(128, 128, 1), # 91 + 32 = 123 + conv_dw(128, 128, 1), # 123 + 32 = 155 + conv_dw(128, 128, 1), # 155 + 32 = 187 + conv_dw(128, 128, 1), # 187 + 32 = 219 + ) + self.stage3 = nn.Sequential( + conv_dw(128, 256, 2), # 219 +3 2 = 241 + conv_dw(256, 256, 1), # 241 + 64 = 301 + ) + + def forward(self, x): + x = self.stage1(x) + x1 = self.stage2(x) + x2 = self.stage3(x1) + + return x, x1, x2 \ No newline at end of file diff --git a/models/cv/detection/retinaface/ixrt/models/prior_box.py b/models/cv/detection/retinaface/ixrt/models/prior_box.py new file mode 100644 index 0000000000000000000000000000000000000000..c6a79195ca254e797fe16474286541ba936e383e --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/models/prior_box.py @@ -0,0 +1,49 @@ +# Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import torch +from itertools import product as product +from math import ceil + +from models.config import cfg_mnet + +class PriorBox(object): + def __init__(self, cfg=cfg_mnet, image_size=None): + super(PriorBox, self).__init__() + + self.min_sizes = cfg['min_sizes'] + self.steps = cfg['steps'] + self.clip = cfg['clip'] + self.image_size = image_size + self.feature_maps = [[ceil(self.image_size[0]/step), ceil(self.image_size[1]/step)] for step in self.steps] + + def forward(self): + anchors = [] + for k, f in enumerate(self.feature_maps): + min_sizes = self.min_sizes[k] + for i, j in product(range(f[0]), range(f[1])): + for min_size in min_sizes: + s_kx = min_size / self.image_size[1] + s_ky = min_size / self.image_size[0] + dense_cx = [x * self.steps[k] / self.image_size[1] for x in [j + 0.5]] + dense_cy = [y * self.steps[k] / self.image_size[0] for y in [i + 0.5]] + for cy, cx in product(dense_cy, dense_cx): + anchors += [cx, cy, s_kx, s_ky] + + # back to torch land + output = torch.Tensor(anchors).view(-1, 4) + if self.clip: + output.clamp_(max=1, min=0) + return output \ No newline at end of file diff --git a/models/cv/detection/retinaface/ixrt/models/retinaface.py b/models/cv/detection/retinaface/ixrt/models/retinaface.py new file mode 100644 index 0000000000000000000000000000000000000000..345c76d409a582044065d390c12f2943ff4f17ca --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/models/retinaface.py @@ -0,0 +1,134 @@ +# Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + + +import torch +import torch.nn as nn +import torchvision.models._utils as _utils +import torch.nn.functional as F +from collections import OrderedDict + +from models.net import MobileNetV1, FPN, SSH + +__all__ = [ + "mnetv1_retinaface", +] + +class ClassHead(nn.Module): + def __init__(self,inchannels=512,num_anchors=3): + super(ClassHead,self).__init__() + self.num_anchors = num_anchors + self.conv1x1 = nn.Conv2d(inchannels,self.num_anchors*2,kernel_size=(1,1),stride=1,padding=0) + + def forward(self,x): + out = self.conv1x1(x) + return out + +class BboxHead(nn.Module): + def __init__(self,inchannels=512,num_anchors=3): + super(BboxHead,self).__init__() + self.conv1x1 = nn.Conv2d(inchannels,num_anchors*4,kernel_size=(1,1),stride=1,padding=0) + + def forward(self,x): + out = self.conv1x1(x) + return out + +class LandmarkHead(nn.Module): + def __init__(self,inchannels=512,num_anchors=3): + super(LandmarkHead,self).__init__() + self.conv1x1 = nn.Conv2d(inchannels,num_anchors*10,kernel_size=(1,1),stride=1,padding=0) + + def forward(self,x): + out = self.conv1x1(x) + return out + +class RetinaFace(nn.Module): + def __init__(self, backbone): + super(RetinaFace,self).__init__() + self.body = backbone + + in_channels_stage2 = 32 + in_channels_list = [ + in_channels_stage2 * 2, + in_channels_stage2 * 4, + in_channels_stage2 * 8, + ] + + out_channels = 64 + self.fpn = FPN(in_channels_list, out_channels) + self.ssh1 = SSH(out_channels, out_channels) + self.ssh2 = SSH(out_channels, out_channels) + self.ssh3 = SSH(out_channels, out_channels) + + self.ClassHead = self._make_class_head(fpn_num=3, inchannels=out_channels) + self.BboxHead = self._make_bbox_head(fpn_num=3, inchannels=out_channels) + self.LandmarkHead = self._make_landmark_head(fpn_num=3, inchannels=out_channels) + + def _make_class_head(self, fpn_num=3, inchannels=64, anchor_num=2): + classhead = nn.ModuleList() + for i in range(fpn_num): + classhead.append(ClassHead(inchannels, anchor_num)) + return classhead + + def _make_bbox_head(self, fpn_num=3, inchannels=64, anchor_num=2): + bboxhead = nn.ModuleList() + for i in range(fpn_num): + bboxhead.append(BboxHead(inchannels, anchor_num)) + return bboxhead + + def _make_landmark_head(self, fpn_num=3, inchannels=64, anchor_num=2): + landmarkhead = nn.ModuleList() + for i in range(fpn_num): + landmarkhead.append(LandmarkHead(inchannels, anchor_num)) + return landmarkhead + + def forward(self, inputs): + + out = self.body(inputs) + + # FPN + fpn = self.fpn(out) + + # SSH + feature1 = self.ssh1(fpn[0]) + feature2 = self.ssh2(fpn[1]) + feature3 = self.ssh3(fpn[2]) + + bbox_out0 = self.BboxHead[0](feature1) + bbox_out1 = self.BboxHead[1](feature2) + bbox_out2 = self.BboxHead[2](feature3) + + cls_out0 = self.ClassHead[0](feature1) + cls_out1 = self.ClassHead[1](feature2) + cls_out2 = self.ClassHead[2](feature3) + + ldm_out0 = self.LandmarkHead[0](feature1) + ldm_out1 = self.LandmarkHead[1](feature2) + ldm_out2 = self.LandmarkHead[2](feature3) + + return bbox_out0, cls_out0, ldm_out0, \ + bbox_out1, cls_out1, ldm_out1, \ + bbox_out2, cls_out2, ldm_out2 + + +def get_model(name): + model = globals()[name]() + return model + +def mnetv1_retinaface(): + backbone = MobileNetV1() + model = RetinaFace(backbone) + + return model \ No newline at end of file diff --git a/models/cv/detection/retinaface/ixrt/requirements.txt b/models/cv/detection/retinaface/ixrt/requirements.txt new file mode 100644 index 0000000000000000000000000000000000000000..c66ca834d47d9c2ab2fe227764f3e43ace932dc5 --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/requirements.txt @@ -0,0 +1,6 @@ +tqdm +tabulate +cuda-python +onnx +onnxsim +opencv-python==4.6.0.66 \ No newline at end of file diff --git a/models/cv/detection/retinaface/ixrt/scripts/infer_retinaface_fp16_accuracy.sh b/models/cv/detection/retinaface/ixrt/scripts/infer_retinaface_fp16_accuracy.sh new file mode 100644 index 0000000000000000000000000000000000000000..5722980c09728dc0e0fc0bd131bc51c126fa0290 --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/scripts/infer_retinaface_fp16_accuracy.sh @@ -0,0 +1,125 @@ +#!/bin/bash +# Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +EXIT_STATUS=0 +check_status() +{ + if ((${PIPESTATUS[0]} != 0));then + EXIT_STATUS=1 + fi +} + +# Run paraments +BSZ=32 +TGT=-1 +WARM_UP=0 +LOOP_COUNT=-1 +RUN_MODE=ACC +PRECISION=float16 + +# Update arguments +index=0 +options=$@ +arguments=($options) +for argument in $options +do + index=`expr $index + 1` + case $argument in + --bs) BSZ=${arguments[index]};; + --tgt) TGT=${arguments[index]};; + esac +done + +PROJ_DIR=${PROJ_DIR:-"."} +DATASETS_DIR="${DATASETS_DIR}" +CHECKPOINTS_DIR="${PROJ_DIR}" +RUN_DIR="${PROJ_DIR}" +CONFIG_DIR="${PROJ_DIR}/config/RETINAFACE_CONFIG" +source ${CONFIG_DIR} +ORIGINE_MODEL=${CHECKPOINTS_DIR}/${ORIGINE_MODEL} + +echo CHECKPOINTS_DIR : ${CHECKPOINTS_DIR} +echo DATASETS_DIR : ${DATASETS_DIR} +echo RUN_DIR : ${RUN_DIR} +echo CONFIG_DIR : ${CONFIG_DIR} +echo ====================== Model Info ====================== +echo Model Name : ${MODEL_NAME} +echo Onnx Path : ${ORIGINE_MODEL} +SIM_MODEL=${ORIGINE_MODEL} + +step=0 + +# Quant Model +if [ $PRECISION == "int8" ];then + let step++ + echo; + echo [STEP ${step}] : Quant Model + if [[ -z ${QUANT_EXIST_ONNX} ]];then + QUANT_EXIST_ONNX=$CHECKPOINTS_DIR/quantized_${MODEL_NAME}.onnx + fi + if [[ -f ${QUANT_EXIST_ONNX} ]];then + SIM_MODEL=${QUANT_EXIST_ONNX} + echo " "Quant Model Skip, ${QUANT_EXIST_ONNX} has been existed + else + python3 ${RUN_DIR}/quant.py \ + --model ${SIM_MODEL} \ + --model_name ${MODEL_NAME} \ + --dataset_dir ${DATASETS_DIR} \ + --observer ${QUANT_OBSERVER} \ + --disable_quant_names ${DISABLE_QUANT_LIST[@]} \ + --save_dir $CHECKPOINTS_DIR \ + --bsz ${QUANT_BATCHSIZE} \ + --step ${QUANT_STEP} \ + --seed ${QUANT_SEED} \ + --imgsz ${IMGSIZE} + SIM_MODEL=${QUANT_EXIST_ONNX} + echo " "Generate ${SIM_MODEL} + fi +fi + +# Build Engine +let step++ +echo; +echo [STEP ${step}] : Build Engine +ENGINE_FILE=${CHECKPOINTS_DIR}/${MODEL_NAME}_${PRECISION}_bs${BSZ}.engine +if [ -f $ENGINE_FILE ];then + echo " "Build Engine Skip, $ENGINE_FILE has been existed +else + python3 ${RUN_DIR}/build_engine.py \ + --precision ${PRECISION} \ + --model ${SIM_MODEL} \ + --engine ${ENGINE_FILE} + echo " "Generate Engine ${ENGINE_FILE} +fi + +# Inference +let step++ +echo; +echo [STEP ${step}] : Inference +python3 ${RUN_DIR}/inference.py \ + --engine_file=${ENGINE_FILE} \ + --datasets_dir=${DATASETS_DIR} \ + --imgsz=${IMGSIZE} \ + --save_folder="${CHECKPOINTS_DIR}/widerface_txt_${BSZ}" \ + --pred_path="${CHECKPOINTS_DIR}/widerface_txt_${BSZ}" \ + --gt=${GT_DIR} \ + --warm_up=${WARM_UP} \ + --loop_count ${LOOP_COUNT} \ + --test_mode ${RUN_MODE} \ + --acc_target ${TGT} \ + --bsz ${BSZ}; check_status + +exit ${EXIT_STATUS} \ No newline at end of file diff --git a/models/cv/detection/retinaface/ixrt/scripts/infer_retinaface_fp16_performance.sh b/models/cv/detection/retinaface/ixrt/scripts/infer_retinaface_fp16_performance.sh new file mode 100644 index 0000000000000000000000000000000000000000..ddbcc65577d11410149adeb1c08cdbb305651ce5 --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/scripts/infer_retinaface_fp16_performance.sh @@ -0,0 +1,125 @@ +#!/bin/bash +# Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +EXIT_STATUS=0 +check_status() +{ + if ((${PIPESTATUS[0]} != 0));then + EXIT_STATUS=1 + fi +} + +# Run paraments +BSZ=32 +TGT=-1 +WARM_UP=3 +LOOP_COUNT=20 +RUN_MODE=FPS +PRECISION=float16 + +# Update arguments +index=0 +options=$@ +arguments=($options) +for argument in $options +do + index=`expr $index + 1` + case $argument in + --bs) BSZ=${arguments[index]};; + --tgt) TGT=${arguments[index]};; + esac +done + +PROJ_DIR=${PROJ_DIR:-"."} +DATASETS_DIR="${DATASETS_DIR}" +CHECKPOINTS_DIR="${PROJ_DIR}" +RUN_DIR="${PROJ_DIR}" +CONFIG_DIR="${PROJ_DIR}/config/RETINAFACE_CONFIG" +source ${CONFIG_DIR} +ORIGINE_MODEL=${CHECKPOINTS_DIR}/${ORIGINE_MODEL} + +echo CHECKPOINTS_DIR : ${CHECKPOINTS_DIR} +echo DATASETS_DIR : ${DATASETS_DIR} +echo RUN_DIR : ${RUN_DIR} +echo CONFIG_DIR : ${CONFIG_DIR} +echo ====================== Model Info ====================== +echo Model Name : ${MODEL_NAME} +echo Onnx Path : ${ORIGINE_MODEL} +SIM_MODEL=${ORIGINE_MODEL} + +step=0 + +# Quant Model +if [ $PRECISION == "int8" ];then + let step++ + echo; + echo [STEP ${step}] : Quant Model + if [[ -z ${QUANT_EXIST_ONNX} ]];then + QUANT_EXIST_ONNX=$CHECKPOINTS_DIR/quantized_${MODEL_NAME}.onnx + fi + if [[ -f ${QUANT_EXIST_ONNX} ]];then + SIM_MODEL=${QUANT_EXIST_ONNX} + echo " "Quant Model Skip, ${QUANT_EXIST_ONNX} has been existed + else + python3 ${RUN_DIR}/quant.py \ + --model ${SIM_MODEL} \ + --model_name ${MODEL_NAME} \ + --dataset_dir ${DATASETS_DIR} \ + --observer ${QUANT_OBSERVER} \ + --disable_quant_names ${DISABLE_QUANT_LIST[@]} \ + --save_dir $CHECKPOINTS_DIR \ + --bsz ${QUANT_BATCHSIZE} \ + --step ${QUANT_STEP} \ + --seed ${QUANT_SEED} \ + --imgsz ${IMGSIZE} + SIM_MODEL=${QUANT_EXIST_ONNX} + echo " "Generate ${SIM_MODEL} + fi +fi + +# Build Engine +let step++ +echo; +echo [STEP ${step}] : Build Engine +ENGINE_FILE=${CHECKPOINTS_DIR}/${MODEL_NAME}_${PRECISION}_bs${BSZ}.engine +if [ -f $ENGINE_FILE ];then + echo " "Build Engine Skip, $ENGINE_FILE has been existed +else + python3 ${RUN_DIR}/build_engine.py \ + --precision ${PRECISION} \ + --model ${SIM_MODEL} \ + --engine ${ENGINE_FILE} + echo " "Generate Engine ${ENGINE_FILE} +fi + +# Inference +let step++ +echo; +echo [STEP ${step}] : Inference +python3 ${RUN_DIR}/inference.py \ + --engine_file=${ENGINE_FILE} \ + --datasets_dir=${DATASETS_DIR} \ + --imgsz=${IMGSIZE} \ + --save_folder="${CHECKPOINTS_DIR}/widerface_txt_${BSZ}" \ + --pred_path="${CHECKPOINTS_DIR}/widerface_txt_${BSZ}" \ + --gt=${GT_DIR} \ + --warm_up=${WARM_UP} \ + --loop_count ${LOOP_COUNT} \ + --test_mode ${RUN_MODE} \ + --fps_target ${TGT} \ + --bsz ${BSZ}; check_status + +exit ${EXIT_STATUS} \ No newline at end of file diff --git a/models/cv/detection/retinaface/ixrt/setup.py b/models/cv/detection/retinaface/ixrt/setup.py new file mode 100644 index 0000000000000000000000000000000000000000..f9b9897f2abe1b461888dfce285ba16d5ddf3b74 --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/setup.py @@ -0,0 +1,13 @@ +""" +WiderFace evaluation code +author: wondervictor +mail: tianhengcheng@gmail.com +copyright@wondervictor +""" + +from distutils.core import setup, Extension +from Cython.Build import cythonize +import numpy + +package = Extension('bbox', ['box_overlaps.pyx'], include_dirs=[numpy.get_include()]) +setup(ext_modules=cythonize([package])) \ No newline at end of file diff --git a/models/cv/detection/retinaface/ixrt/torch2onnx.py b/models/cv/detection/retinaface/ixrt/torch2onnx.py new file mode 100644 index 0000000000000000000000000000000000000000..747c770aef6072a32bdd602da280719fb3d45b54 --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/torch2onnx.py @@ -0,0 +1,70 @@ +# Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import argparse +import torch +import onnx +from onnxsim import simplify + + +from models import mnetv1_retinaface + +def parse_config(): + parser = argparse.ArgumentParser() + + parser.add_argument( + "--model", + type=str, + help="torch model path", + ) + parser.add_argument( + "--onnx_model", + type=str, + help="onnx model path", + ) + args = parser.parse_args() + return args + + +if __name__ == "__main__": + args = parse_config() + + torch_pretrained = args.model + export_onnx_file = args.onnx_model + + model = mnetv1_retinaface() + model.eval() + + checkpoint = torch.load(torch_pretrained, map_location=torch.device("cpu")) + model.load_state_dict(checkpoint, strict=False) + + inputs = torch.randn(32, 3, 320, 320) + + torch.onnx.export(model, + inputs, + export_onnx_file, + opset_version=11, + input_names=["input"], + output_names = ["bbox_out0", "cls_out0", "ldm_out0", + "bbox_out1", "cls_out1", "ldm_out1", + "bbox_out2", "cls_out2", "ldm_out2"] + ) + + onnx_model = onnx.load(export_onnx_file) # load onnx model + model_simp, check = simplify(onnx_model) + assert check, "Simplified ONNX model could not be validated" + onnx.save(model_simp, export_onnx_file) + print('finished exporting onnx') \ No newline at end of file diff --git a/models/cv/detection/retinaface/ixrt/utils/__init__.py b/models/cv/detection/retinaface/ixrt/utils/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/models/cv/detection/retinaface/ixrt/utils/box_utils.py b/models/cv/detection/retinaface/ixrt/utils/box_utils.py new file mode 100644 index 0000000000000000000000000000000000000000..044d09728bf67814decceada5822478c366bf9ee --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/utils/box_utils.py @@ -0,0 +1,339 @@ +# Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import torch +import numpy as np + + +def point_form(boxes): + """ Convert prior_boxes to (xmin, ymin, xmax, ymax) + representation for comparison to point form ground truth data. + Args: + boxes: (tensor) center-size default boxes from priorbox layers. + Return: + boxes: (tensor) Converted xmin, ymin, xmax, ymax form of boxes. + """ + return torch.cat((boxes[:, :2] - boxes[:, 2:]/2, # xmin, ymin + boxes[:, :2] + boxes[:, 2:]/2), 1) # xmax, ymax + + +def center_size(boxes): + """ Convert prior_boxes to (cx, cy, w, h) + representation for comparison to center-size form ground truth data. + Args: + boxes: (tensor) point_form boxes + Return: + boxes: (tensor) Converted xmin, ymin, xmax, ymax form of boxes. + """ + return torch.cat((boxes[:, 2:] + boxes[:, :2])/2, # cx, cy + boxes[:, 2:] - boxes[:, :2], 1) # w, h + + +def intersect(box_a, box_b): + """ We resize both tensors to [A,B,2] without new malloc: + [A,2] -> [A,1,2] -> [A,B,2] + [B,2] -> [1,B,2] -> [A,B,2] + Then we compute the area of intersect between box_a and box_b. + Args: + box_a: (tensor) bounding boxes, Shape: [A,4]. + box_b: (tensor) bounding boxes, Shape: [B,4]. + Return: + (tensor) intersection area, Shape: [A,B]. + """ + A = box_a.size(0) + B = box_b.size(0) + max_xy = torch.min(box_a[:, 2:].unsqueeze(1).expand(A, B, 2), + box_b[:, 2:].unsqueeze(0).expand(A, B, 2)) + min_xy = torch.max(box_a[:, :2].unsqueeze(1).expand(A, B, 2), + box_b[:, :2].unsqueeze(0).expand(A, B, 2)) + inter = torch.clamp((max_xy - min_xy), min=0) + return inter[:, :, 0] * inter[:, :, 1] + + +def jaccard(box_a, box_b): + """Compute the jaccard overlap of two sets of boxes. The jaccard overlap + is simply the intersection over union of two boxes. Here we operate on + ground truth boxes and default boxes. + E.g.: + A ∩ B / A ∪ B = A ∩ B / (area(A) + area(B) - A ∩ B) + Args: + box_a: (tensor) Ground truth bounding boxes, Shape: [num_objects,4] + box_b: (tensor) Prior boxes from priorbox layers, Shape: [num_priors,4] + Return: + jaccard overlap: (tensor) Shape: [box_a.size(0), box_b.size(0)] + """ + inter = intersect(box_a, box_b) + area_a = ((box_a[:, 2]-box_a[:, 0]) * + (box_a[:, 3]-box_a[:, 1])).unsqueeze(1).expand_as(inter) # [A,B] + area_b = ((box_b[:, 2]-box_b[:, 0]) * + (box_b[:, 3]-box_b[:, 1])).unsqueeze(0).expand_as(inter) # [A,B] + union = area_a + area_b - inter + return inter / union # [A,B] + + +def matrix_iou(a, b): + """ + return iou of a and b, numpy version for data augenmentation + """ + lt = np.maximum(a[:, np.newaxis, :2], b[:, :2]) + rb = np.minimum(a[:, np.newaxis, 2:], b[:, 2:]) + + area_i = np.prod(rb - lt, axis=2) * (lt < rb).all(axis=2) + area_a = np.prod(a[:, 2:] - a[:, :2], axis=1) + area_b = np.prod(b[:, 2:] - b[:, :2], axis=1) + return area_i / (area_a[:, np.newaxis] + area_b - area_i) + + +def matrix_iof(a, b): + """ + return iof of a and b, numpy version for data augenmentation + """ + lt = np.maximum(a[:, np.newaxis, :2], b[:, :2]) + rb = np.minimum(a[:, np.newaxis, 2:], b[:, 2:]) + + area_i = np.prod(rb - lt, axis=2) * (lt < rb).all(axis=2) + area_a = np.prod(a[:, 2:] - a[:, :2], axis=1) + return area_i / np.maximum(area_a[:, np.newaxis], 1) + + +def match(threshold, truths, priors, variances, labels, landms, loc_t, conf_t, landm_t, idx): + """Match each prior box with the ground truth box of the highest jaccard + overlap, encode the bounding boxes, then return the matched indices + corresponding to both confidence and location preds. + Args: + threshold: (float) The overlap threshold used when mathing boxes. + truths: (tensor) Ground truth boxes, Shape: [num_obj, 4]. + priors: (tensor) Prior boxes from priorbox layers, Shape: [n_priors,4]. + variances: (tensor) Variances corresponding to each prior coord, + Shape: [num_priors, 4]. + labels: (tensor) All the class labels for the image, Shape: [num_obj]. + landms: (tensor) Ground truth landms, Shape [num_obj, 10]. + loc_t: (tensor) Tensor to be filled w/ endcoded location targets. + conf_t: (tensor) Tensor to be filled w/ matched indices for conf preds. + landm_t: (tensor) Tensor to be filled w/ endcoded landm targets. + idx: (int) current batch index + Return: + The matched indices corresponding to 1)location 2)confidence 3)landm preds. + """ + # jaccard index + overlaps = jaccard( + truths, + point_form(priors) + ) + # (Bipartite Matching) + # [1,num_objects] best prior for each ground truth + best_prior_overlap, best_prior_idx = overlaps.max(1, keepdim=True) + + # ignore hard gt + valid_gt_idx = best_prior_overlap[:, 0] >= 0.2 + best_prior_idx_filter = best_prior_idx[valid_gt_idx, :] + if best_prior_idx_filter.shape[0] <= 0: + loc_t[idx] = 0 + conf_t[idx] = 0 + return + + # [1,num_priors] best ground truth for each prior + best_truth_overlap, best_truth_idx = overlaps.max(0, keepdim=True) + best_truth_idx.squeeze_(0) + best_truth_overlap.squeeze_(0) + best_prior_idx.squeeze_(1) + best_prior_idx_filter.squeeze_(1) + best_prior_overlap.squeeze_(1) + best_truth_overlap.index_fill_(0, best_prior_idx_filter, 2) # ensure best prior + # TODO refactor: index best_prior_idx with long tensor + # ensure every gt matches with its prior of max overlap + for j in range(best_prior_idx.size(0)): # 判别此anchor是预测哪一个boxes + best_truth_idx[best_prior_idx[j]] = j + matches = truths[best_truth_idx] # Shape: [num_priors,4] 此处为每一个anchor对应的bbox取出来 + conf = labels[best_truth_idx] # Shape: [num_priors] 此处为每一个anchor对应的label取出来 + conf[best_truth_overlap < threshold] = 0 # label as background overlap<0.35的全部作为负样本 + loc = encode(matches, priors, variances) + + matches_landm = landms[best_truth_idx] + landm = encode_landm(matches_landm, priors, variances) + loc_t[idx] = loc # [num_priors,4] encoded offsets to learn + conf_t[idx] = conf # [num_priors] top class label for each prior + landm_t[idx] = landm + + +def encode(matched, priors, variances): + """Encode the variances from the priorbox layers into the ground truth boxes + we have matched (based on jaccard overlap) with the prior boxes. + Args: + matched: (tensor) Coords of ground truth for each prior in point-form + Shape: [num_priors, 4]. + priors: (tensor) Prior boxes in center-offset form + Shape: [num_priors,4]. + variances: (list[float]) Variances of priorboxes + Return: + encoded boxes (tensor), Shape: [num_priors, 4] + """ + + # dist b/t match center and prior's center + g_cxcy = (matched[:, :2] + matched[:, 2:])/2 - priors[:, :2] + # encode variance + g_cxcy /= (variances[0] * priors[:, 2:]) + # match wh / prior wh + g_wh = (matched[:, 2:] - matched[:, :2]) / priors[:, 2:] + g_wh = torch.log(g_wh) / variances[1] + # return target for smooth_l1_loss + return torch.cat([g_cxcy, g_wh], 1) # [num_priors,4] + +def encode_landm(matched, priors, variances): + """Encode the variances from the priorbox layers into the ground truth boxes + we have matched (based on jaccard overlap) with the prior boxes. + Args: + matched: (tensor) Coords of ground truth for each prior in point-form + Shape: [num_priors, 10]. + priors: (tensor) Prior boxes in center-offset form + Shape: [num_priors,4]. + variances: (list[float]) Variances of priorboxes + Return: + encoded landm (tensor), Shape: [num_priors, 10] + """ + + # dist b/t match center and prior's center + matched = torch.reshape(matched, (matched.size(0), 5, 2)) + priors_cx = priors[:, 0].unsqueeze(1).expand(matched.size(0), 5).unsqueeze(2) + priors_cy = priors[:, 1].unsqueeze(1).expand(matched.size(0), 5).unsqueeze(2) + priors_w = priors[:, 2].unsqueeze(1).expand(matched.size(0), 5).unsqueeze(2) + priors_h = priors[:, 3].unsqueeze(1).expand(matched.size(0), 5).unsqueeze(2) + priors = torch.cat([priors_cx, priors_cy, priors_w, priors_h], dim=2) + g_cxcy = matched[:, :, :2] - priors[:, :, :2] + # encode variance + g_cxcy /= (variances[0] * priors[:, :, 2:]) + # g_cxcy /= priors[:, :, 2:] + g_cxcy = g_cxcy.reshape(g_cxcy.size(0), -1) + # return target for smooth_l1_loss + return g_cxcy + + +def decode(loc, priors, variances): + """Decode locations from predictions using priors to undo + the encoding we did for offset regression at train time. + Args: + loc (tensor): location predictions for loc layers, + Shape: [num_priors,4] + priors (tensor): Prior boxes in center-offset form. + Shape: [num_priors,4]. + variances: (list[float]) Variances of priorboxes + Return: + decoded bounding box predictions + """ + + boxes = torch.cat(( + priors[:, :2] + loc[:, :2] * variances[0] * priors[:, 2:], + priors[:, 2:] * torch.exp(loc[:, 2:] * variances[1])), 1) + boxes[:, :2] -= boxes[:, 2:] / 2 + boxes[:, 2:] += boxes[:, :2] + return boxes + +def decode_landm(pre, priors, variances): + """Decode landm from predictions using priors to undo + the encoding we did for offset regression at train time. + Args: + pre (tensor): landm predictions for loc layers, + Shape: [num_priors,10] + priors (tensor): Prior boxes in center-offset form. + Shape: [num_priors,4]. + variances: (list[float]) Variances of priorboxes + Return: + decoded landm predictions + """ + landms = torch.cat((priors[:, :2] + pre[:, :2] * variances[0] * priors[:, 2:], + priors[:, :2] + pre[:, 2:4] * variances[0] * priors[:, 2:], + priors[:, :2] + pre[:, 4:6] * variances[0] * priors[:, 2:], + priors[:, :2] + pre[:, 6:8] * variances[0] * priors[:, 2:], + priors[:, :2] + pre[:, 8:10] * variances[0] * priors[:, 2:], + ), dim=1) + return landms + + +def log_sum_exp(x): + """Utility function for computing log_sum_exp while determining + This will be used to determine unaveraged confidence loss across + all examples in a batch. + Args: + x (Variable(tensor)): conf_preds from conf layers + """ + x_max = x.data.max() + return torch.log(torch.sum(torch.exp(x-x_max), 1, keepdim=True)) + x_max + + +def nms(boxes, scores, overlap=0.5, top_k=200): + """Apply non-maximum suppression at test time to avoid detecting too many + overlapping bounding boxes for a given object. + Args: + boxes: (tensor) The location preds for the img, Shape: [num_priors,4]. + scores: (tensor) The class predscores for the img, Shape:[num_priors]. + overlap: (float) The overlap thresh for suppressing unnecessary boxes. + top_k: (int) The Maximum number of box preds to consider. + Return: + The indices of the kept boxes with respect to num_priors. + """ + + keep = torch.Tensor(scores.size(0)).fill_(0).long() + if boxes.numel() == 0: + return keep + x1 = boxes[:, 0] + y1 = boxes[:, 1] + x2 = boxes[:, 2] + y2 = boxes[:, 3] + area = torch.mul(x2 - x1, y2 - y1) + v, idx = scores.sort(0) # sort in ascending order + # I = I[v >= 0.01] + idx = idx[-top_k:] # indices of the top-k largest vals + xx1 = boxes.new() + yy1 = boxes.new() + xx2 = boxes.new() + yy2 = boxes.new() + w = boxes.new() + h = boxes.new() + + # keep = torch.Tensor() + count = 0 + while idx.numel() > 0: + i = idx[-1] # index of current largest val + # keep.append(i) + keep[count] = i + count += 1 + if idx.size(0) == 1: + break + idx = idx[:-1] # remove kept element from view + # load bboxes of next highest vals + torch.index_select(x1, 0, idx, out=xx1) + torch.index_select(y1, 0, idx, out=yy1) + torch.index_select(x2, 0, idx, out=xx2) + torch.index_select(y2, 0, idx, out=yy2) + # store element-wise max with next highest score + xx1 = torch.clamp(xx1, min=x1[i]) + yy1 = torch.clamp(yy1, min=y1[i]) + xx2 = torch.clamp(xx2, max=x2[i]) + yy2 = torch.clamp(yy2, max=y2[i]) + w.resize_as_(xx2) + h.resize_as_(yy2) + w = xx2 - xx1 + h = yy2 - yy1 + # check sizes of xx1 and xx2.. after each iteration + w = torch.clamp(w, min=0.0) + h = torch.clamp(h, min=0.0) + inter = w*h + # IoU = i / (area(a) + area(b) - i) + rem_areas = torch.index_select(area, 0, idx) # load remaining areas) + union = (rem_areas - inter) + area[i] + IoU = inter/union # store result in iou + # keep only elements with an IoU <= overlap + idx = idx[IoU.le(overlap)] + return keep, count \ No newline at end of file diff --git a/models/cv/detection/retinaface/ixrt/utils/nms/py_cpu_nms.py b/models/cv/detection/retinaface/ixrt/utils/nms/py_cpu_nms.py new file mode 100644 index 0000000000000000000000000000000000000000..6991f859bf506f937c1bdbb358f5eab31d3cd1d7 --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/utils/nms/py_cpu_nms.py @@ -0,0 +1,46 @@ +# Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import numpy as np + +def py_cpu_nms(dets, thresh): + """Pure Python NMS baseline.""" + x1 = dets[:, 0] + y1 = dets[:, 1] + x2 = dets[:, 2] + y2 = dets[:, 3] + scores = dets[:, 4] + + areas = (x2 - x1 + 1) * (y2 - y1 + 1) + order = scores.argsort()[::-1] + + keep = [] + while order.size > 0: + i = order[0] + keep.append(i) + xx1 = np.maximum(x1[i], x1[order[1:]]) + yy1 = np.maximum(y1[i], y1[order[1:]]) + xx2 = np.minimum(x2[i], x2[order[1:]]) + yy2 = np.minimum(y2[i], y2[order[1:]]) + + w = np.maximum(0.0, xx2 - xx1 + 1) + h = np.maximum(0.0, yy2 - yy1 + 1) + inter = w * h + ovr = inter / (areas[i] + areas[order[1:]] - inter) + + inds = np.where(ovr <= thresh)[0] + order = order[inds + 1] + + return keep \ No newline at end of file diff --git a/models/cv/detection/retinaface/ixrt/wider_face_dataset.py b/models/cv/detection/retinaface/ixrt/wider_face_dataset.py new file mode 100644 index 0000000000000000000000000000000000000000..8095cb6f05897c19d387230755b32c1ae6ad3352 --- /dev/null +++ b/models/cv/detection/retinaface/ixrt/wider_face_dataset.py @@ -0,0 +1,81 @@ +# Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +import os +import torch +import torch.utils.data as data +import cv2 +import numpy as np + +def lt_preproc(img, input_size=(320, 320), swap=(2, 0, 1), mean=(104.0, 117.0, 123.0), std=1.0): + if len(img.shape) == 3: + padded_img = np.ones((input_size[0], input_size[1], 3), dtype=np.float32) * 114.0 + else: + padded_img = np.ones(input_size, dtype=np.float32) * 114.0 + + r = min(input_size[0] / img.shape[0], input_size[1] / img.shape[1]) + resized_img = cv2.resize( + img, + (int(img.shape[1] * r), int(img.shape[0] * r)), + interpolation=cv2.INTER_LINEAR, + ) + + padded_img[: int(img.shape[0] * r), : int(img.shape[1] * r)] = resized_img + + padded_img -= mean + padded_img /= std + + if swap is not None: + padded_img = padded_img.transpose(swap) + padded_img = np.ascontiguousarray(padded_img, dtype=np.float32) + + return padded_img, r + + +class WiderFaceDetection(data.Dataset): + def __init__(self, prj_dir, preproc=lt_preproc, input_size=(320, 320)): + self.preproc = preproc + self.input_size = input_size + self.image_dir = os.path.join(prj_dir, "images") + + testset_list = os.path.join(prj_dir, "wider_val.txt") + with open(testset_list, 'r') as fr: + self.imgs_path = fr.read().split() + + self.count = 0 + + def __len__(self): + return len(self.imgs_path) + + def __getitem__(self, index): + img_file = os.path.join(self.image_dir, self.imgs_path[index]) + img = cv2.imread(img_file, cv2.IMREAD_COLOR) + + if self.preproc is not None: + img, r = self.preproc(img, self.input_size) + + return torch.from_numpy(img), r, self.imgs_path[index] + + +def detection_collate(batch): + imgs, rs, img_files = list(), list(), list() + for i, (img, scale, img_file) in enumerate(batch): + if torch.is_tensor(img): + imgs.append(img) + rs.append(scale) + img_files.append(img_file) + + imgs = torch.stack(imgs, 0) + return imgs, torch.from_numpy(np.array(rs)), img_files \ No newline at end of file