# pytorch_modelzoo

**Repository Path**: cambricon/pytorch_modelzoo

## Basic Information

- **Project Name**: pytorch_modelzoo
- **Description**: cambriocn pytorch训练和推理模型集合
- **Primary Language**: Unknown
- **License**: Apache-2.0
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 58
- **Forks**: 30
- **Created**: 2022-10-14
- **Last Updated**: 2024-09-02

## Categories & Tags

**Categories**: ai

**Tags**: None

## README

# PyTorch ModelZoo 

## 介绍

PyTorch是时下最流行的AI框架，寒武纪对其进行了定制化开发，新增了对寒武纪加速板卡及寒武纪AI软件栈的支持，通常称之为Cambricon PyTorch。相比于原生PyTorch，用户基本不用做任何代码改动即可快速地将AI模型迁移至Cambricon PyTorch上。

针对CV 分类、检测、分割、NLP、语音等场景常用的各类经典和前沿的AI模型，本仓库展示了如何对其进行适配，使其可运行在Cambricon PyTorch上。开发者在进行其他AI 应用移植时可参考本仓库。


## 网络支持列表和链接

CV：

| MODELS | FRAMEWORK | Train Mode |Distributed Train| Infer  Mode
| ------------- | ------------- | ------------- | ------------- | ------------- | 
| [ResNet50](built-in/cv/classification/vision_classification) | PyTorch1.6|FP32&&AMP|YES| CNNL |
| [ResNet18](built-in/cv/classification/vision_classification) | PyTorch1.6|FP32|YES| CNNL | 
| [VGG16](built-in/cv/classification/vision_classification) | PyTorch1.6|FP32&&AMP|YES| CNNL | 
| [MobileNetv2](built-in/cv/classification/vision_classification) | PyTorch1.6|FP32&&AMP|YES| CNNL | 
| [AlexNet](built-in/cv/classification/vision_classification) | PyTorch1.6|FP32|YES| CNNL | 
| [GoogLeNet](built-in/cv/classification/vision_classification) | PyTorch1.6|FP32&&AMP|YES| CNNL | 
| [ResNet101](built-in/cv/classification/vision_classification) | PyTorch1.6|FP32|YES| CNNL | 
| [VGG19](built-in/cv/classification/vision_classification) | PyTorch1.6|FP32|YES| CNNL | 
| [VGG16_bn](built-in/cv/classification/vision_classification) | PyTorch1.6|FP32|YES| CNNL |
| [ShuffleNet_v2_x0_5](built-in/cv/classification/vision_classification) | PyTorch1.6|FP32|YES| CNNL |
| [ShuffleNet_v2_x1_0](built-in/cv/classification/vision_classification) | PyTorch1.6|FP32|YES| CNNL |
| [ShuffleNet_v2_x1_5](built-in/cv/classification/vision_classification) | PyTorch1.6|FP32|YES| CNNL |
| [ngc-resnet50v1_5](built-in/cv/classification/ngc-resnet50v1_5) | PyTorch1.6|FP32&&AMP|YES| CNNL |
| [Inceptionv2](built-in/cv/classification/Inceptionv2) | PyTorch1.6|FP32&&AMP|YES| CNNL |
| [Inceptionv3](built-in/cv/classification/timm) | PyTorch1.6|FP32&&AMP|YES| CNNL |
| [Inceptionv4](built-in/cv/classification/timm) | PyTorch1.9|FP32&&AMP|YES| CNNL |
| [OLTR](built-in/cv/classification/OLTR) | PyTorch1.6|FP32|NO| CNNL |
| [P3D](built-in/cv/classification/P3D) | PyTorch1.6|FP32|FP32&&AMP| CNNL |
| [Swin-Transformer-SSL](built-in/cv/classification/Swin-Transformer-SSL) | PyTorch1.9|FP32&&AMP|YES| CNNL |
| [swin_transformer](built-in/cv/classification/swin_transformer) | PyTorch1.6|AMP|YES| CNNL |
| [crnn](built-in/cv/classification/crnn) | PyTorch1.6|FP32|YES| CNNL |
| [FasterRCNN](built-in/cv/detection/FasterRCNN) | PyTorch1.6|FP32|YES| CNNL |
| [MaskRCNN](built-in/cv/detection/MaskRCNN) | PyTorch1.6|FP32|YES| CNNL |
| [RFBNet](built-in/cv/detection/RFBNet) | PyTorch1.6|FP32|FP32&&AMP| CNNL |
| [SSD_ResNet50](built-in/cv/detection/SSD_ResNet50) | PyTorch1.6|FP32&&AMP|YES| CNNL |
| [SSD_VGG16](built-in/cv/detection/SSD_VGG16) | PyTorch1.6|FP32&&AMP|YES| CNNL |
| [PointPillar](built-in/cv/detection/PointPillar) | PyTorch1.9|FP32|YES| CNNL |
| [unet3d](built-in/cv/segmentation/unet3d) | PyTorch1.6|FP32|YES| CNNL |
| [CycleGAN_and_pix2pix](built-in/cv/GAN/CycleGAN_and_pix2pix) | PyTorch1.6|FP32|NO| CNNL |
| [enet](built-in/cv/segmentation/enet) | PyTorch1.9|FP32&&AMP|YES| CNNL |

NLP:

| MODELS | FRAMEWORK | Train Mode |Distributed Train| Infer  Mode
| ------------- | ------------- | ------------- | ------------- | ------------- | 
| [transformer](built-in/nlp/transformer) | PyTorch1.6|FP32|YES| CNNL |
| [BERT](built-in/nlp/BERT/) | PyTorch1.6|FP32|YES| CNNL |
| [bert-base-cased](built-in/nlp/bert-base-cased) | PyTorch1.6|FP32&&AMP|YES| CNNL |
| [CRF](built-in/nlp/CRF) | PyTorch1.6|FP32|NO| CNNL |
| [bert_base_finetune_msra_ner](built-in/nlp/bert_base_finetune_msra_ner) | PyTorch1.6|FP32&&AMP|YES| CNNL |
| [DeepSpeech2](built-in/nlp/DeepSpeech2) | PyTorch1.6|FP32|YES| CNNL |
| [MT5](built-in/nlp/mt5) | PyTorch1.6 && 1.9|FP32&&AMP|YES| CNNL |

SpeechSynthesis/:

| MODELS | FRAMEWORK | Train Mode |Distributed Train| Infer  Mode
| ------------- | ------------- | ------------- | ------------- | ------------- | 
| [Tacotron2](built-in/SpeechSynthesis/Tacotron2andWaveGlow) | PyTorch1.6|FP32|YES| CNNL |
| [WaveGlow](built-in/SpeechSynthesis/Tacotron2andWaveGlow) | PyTorch1.6|AMP|YES| CNNL |
| [WaveRNN](built-in/SpeechSynthesis/WaveRNN) | PyTorch1.6|FP32&&AMP|YES| CNNL |

recommendation/:

| MODELS | FRAMEWORK | Train Mode |Distributed Train| Infer  Mode
| ------------- | ------------- | ------------- | ------------- | ------------- | 
| [DLRM](built-in/recommendation/DLRM) | PyTorch1.6|FP32&&AMP|YES| CNNL |

Speech/:

| MODELS | FRAMEWORK | Train Mode |Distributed Train| Infer  Mode
| ------------- | ------------- | ------------- | ------------- | ------------- | 
| [speechbrain](built-in/Speech/speechbrain) | PyTorch1.9|FP32|YES| CNNL |


## issues/wiki/forum 跳转链接

## contrib 指引和链接

## LICENSE

PyTorch ModelZoo  的 License 具体内容请参见[LICENSE](LICENSE)文件。

## 免责声明

PyTorch ModelZoo 仅提供公共数据集以及预训练模型的下载链接，公共数据集及预训练模型并不属于 PyTorch ModelZoo ，PyTorch ModelZoo  也不对其质量或维护承担责任。请您在使用公共数据集和预训练模型的过程中，确保符合其对应的使用许可。

如果您不希望您的数据集或模型公布在 PyTorch ModelZoo上，或者您希望更新 PyTorch ModelZoo中属于您的数据集或模型，请您通过 Gitee 中提交 issue，您也可以联系ecosystem@cambricon.com告知我们。


## Release Note
### v0.5.0:
- Cambricon Torch支持v1.16.0(Cambricon SDK 1.14)

### v0.4.0:
- Cambricon Torch支持v1.15.0(Cambricon SDK 1.13)

### v0.3.0:
- Cambricon Torch支持v1.14.0(Cambricon SDK 1.12)

### v0.2.0:
- Cambricon Torch支持v1.13.0(Cambricon SDK 1.11)

- CV: 新增enet网络的支持，MaskRCNN支持COCO2017数据集训练

- NLP: 新增MT5网络的支持

- Speech: 新增speechbrain网络的支持

### v0.1.0:
- Cambricon Torch支持v1.11.0

- CV: 新增ResNet18/ResNet50/ResNet101/VGG16/VGG16_bn/VGG19/MobileNetv2/AlexNet/GoogLeNet/ShuffleNet_v2_x0_5/ShuffleNet_v2_x1_0/ShuffleNet_v2_x1_5/ngc-resnet50v1_5/Inceptionv2/Inceptionv3/Inceptionv4/OLTR/P3D/Swin-Transformer-SSL/swin_transformer/crnn/MaskRCNN/FasterRCNN/RFBNet/SSD_ResNet50/SSD_VGG16/PointPillar/unet3d/CycleGan_and_pix2pix网络的支持

- NLP: 新增transformer/BERT/bert-base-cased/CRF/bert_base_finetune_msra_ner/DeepSpeech2网络的支持

- SpeechSynthesis: 新增Tacotron2/WaveGlow/WaveRNN网络的支持

- Recommendation: 新增DLRM网络的支持