【清华大学】【高校贡献】【PyTorch】 RegNetX迁移到Pytorch1.8版本 · Pull Request !568 · Ascend/ModelZoo-PyTorch

### RegNetX

This implements training of RegNetX on the ImageNet dataset, mainly modified from [facebookresearch/pycls](http://https://github.com/facebookresearch/pycls).

### RegNetX Detail

As of the current date, Ascend-Pytorch is still inefficient for contiguous operations.Therefore, RegNetX is re-implemented using semantics such as custom OP.

### Requirements

- Install PyTorch (pytorch.org)
- pip install -r requirements.txt
- Download the ImageNet dataset from [http://www.image-net.org/](http://http://www.image-net.org/). Then, and move validation images to labeled subfolders, using [the following shell script](http://https://raw.githubusercontent.com/soumith/imagenetloader.torch/master/valprep.sh)

### Training

To train a model, run `main.py` with the desired model architecture and the path to the ImageNet dataset:

```
# training 1p performance
bash ./test/train_performance_1p.sh --data_path=real_data_path

# training 8p accuracy
bash ./test/train_full_8p.sh --data_path=real_data_path

# training 8p performance
bash ./test/train_performance_8p.sh --data_path=real_data_path

# 8p eval
bash test/train_eval_8p.sh --data_path=real_data_path --pth_path=./checkpoints/model_best.pth.tar

# finetuning
bash test/train_finetune_1p.sh --data_path=real_data_path --pth_path=./checkpoints/checkpoint.pth.tar

# online inference demo 
python3.7.5 demo.py
```

Log path: test/output/devie_id/train_${device_id}.log # training detail log test/output/devie_id/RegNetX_8p_perf.log # 8p training performance result log test/output/devie_id/RegNetX_8p_acc.log # 8p training accuracy result log

### RegNetX training result
| Acc@1  | FPS          | Pytorch_version | Npu_nums | Epochs | AMP_Type |
|--------|--------------|-----------------|----------|--------|----------|
| -      | 1649         | 1.5             | 1        | 1      | O2       |
| 77.167 | 5900（显示3100） | 1.5             | 8        | 100    | O2       |
| -      | 1282         | 1.8             | 1        | 1      | O2       |
| 77.33  | 4900（显示2750） | 1.8             | 8        | 100    | O2       |

Ascend/ModelZoo-PyTorch .gitee-modal { width: 500px !important; }

搜索帮助

Ascend/ModelZoo-PyTorch