# Iluvatar CoreX IxRT **Repository Path**: wz50525/iluvatar-corex-ixrt ## Basic Information - **Project Name**: Iluvatar CoreX IxRT - **Description**: 本仓库为天数智芯Iluvatar Corex IxRT高性能推理引擎的开源代码部分,提供IxRT相关插件、部署工具和演示IxRT平台用法和能力的应用示例。 - **Primary Language**: Unknown - **License**: Apache-2.0 - **Default Branch**: master - **Homepage**: None - **GVP Project**: No ## Statistics - **Stars**: 0 - **Forks**: 16 - **Created**: 2024-10-17 - **Last Updated**: 2024-10-17 ## Categories & Tags **Categories**: Uncategorized **Tags**: None ## README # IxRT Open Source Software This repository contains the Open Source Software (OSS) components of the Iluvatar Corex IxRT. It includes the sources for IxRT plugins and deploy tools, as well as sample applications demonstrating the usages and capabilities of the IxRT platform. Iluvatar CoreX IxRT is a high-performance AI inference engine. IxRT provides an AI compiler, an inference runtime and development APIs for AI inference applications. Directory structure: ```bash . ├── CMakeLists.txt # To build samples ├── README.md # This readme ├── cmake # Various cmake scripts ├── requirements # Python requirements ├── samples # IxRT samples ├── ixrt # IxRT Python code, include its deploy tool └── tools # IxRT various tools, please check readme for details ``` ## Prebuilt IxRT Python Package We provide the IxRT Python package for an easy installation. To install: ```bash pip3 install ixrt ``` ## Build ### Prerequisites To build the IxRT-OSS components, you will first need some software packages, The required dependencies for installation are located in the folder named requirements. ```bash cd $IxRT_OSS_PATH/requirements pip3 install -r requirements-ixrt.txt pip3 install -r requirements-quant-tests.txt pip3 install -r requirements-quant.txt ``` To build IxRT C++ samples, you will also need to download IxRT `.run` or `.tar.gz` software package from Iluvatar official website, and install it following procedures in **Install** chapter of IxRT documentation. ### Building IxRT-OSS #### Generate Makefiles and build ```bash cd /path/to/ixrt-oss cmake -B build cmake --build build -j ``` ## Samples Lists Below is a sample of the model running with IxRT. | Sample | Language | Format | Description | | ------------------------------------------- | -------- | ------ | ---------------------------------------------------------- | | [sampleYoloV3](samples/sampleYoloV3) | C++ | ONNX | Execute YoloV3 with IxRT API | | [sampleYoloX](samples/sampleYoloX) | C++ | ONNX | Execute YoloX-m with IxRT API | | [sampleYoloV5](samples/sampleYoloV5) | C++ | ONNX | Execute YoloV5-s with IxRT API | | [resnet18](samples/python/resnet18) | Python | ONNX | Execute ResNet18 with IxRT API | | [sampleResNet](samples/sampleResNet) | C++ | ONNX | Execute ResNet18 with IxRT API | | [quantization](samples/python/quantization) | Python | ONNX | Demonstrate how to useIxRT deploy tool to quantize an onnx | ### Advanced samples | Sample | Language | Format | Description | | -------------------------------------------------------- | -------- | ------ | ---------------------------------------------------------------------------------------- | | [sampleHideDataTransfer](samples/sampleHideDataTransfer) | C++ | ONNX | Hide data transfer with ResNet18 ONNX | | [sampleAPIToResnet](samples/sampleAPIToResnet) | C++ | ONNX | Construct a ResNet18 network with IxRT custom layer API | | [sampleAPIToInt8Resnet](samples/sampleAPIToInt8Resnet) | C++ | ONNX | Construct an int8 ResNet18 network with IxRT custom layer API | | [sampleBert](samples/sampleBert) | Python | ONNX | Construct an int8 and fp16 Bert with IxRT custom layer API | | [sampleQuant](samples/sampleQuant) | C++ | ONNX | Demonstrate how to useIxRT deploy tool to quantize an onnx | | [python](samples/python) | Python | ONNX | IXRT Quantization and Example Tool Using Python API | | [plugins](samples/plugins) | C++ | ONNX | The plugin library provided by IxRT, which includes the NMSplugin and YOLOdecoder plugin | ## License Copyright (c) 2024 Iluvatar CoreX. All rights reserved. This project has an Apache-2.0 license, as found in the [LICENSE](LICENSE) file.