VoVNet-39

This implements training of VoVNet-39 on the ImageNet dataset, mainly modified from GitHub.

VoVNet-39 Detail

As of the current date, Ascend-Pytorch is still inefficient for contiguous operations.Therefore, VoVNet-39 is re-implemented using semantics such as custom OP.

Requirements

• Install PyTorch (pytorch.org)
• pip install -r requirements.txt
• torchvision==0.5.0(x86) && torchvision==0.2.0(arm)
• Download the ImageNet dataset from http://www.image-net.org/
  • Then, and move validation images to labeled subfolders, using the following shell script

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

Log path: - test/output/devie_id/train_${device_id}.log # training detail log - test/output/devie_id/VoVNet39_for_PyTorch_bs128_8p_perf.log # 8p training performance result log - test/output/devie_id/VoVNet39_for_PyTorch_bs128_8p_acc.log # 8p training accuracy result log

Running demo

Run demo.py as the demo script with trained model (model.pth as example):

python3 demo.py model.pth

VoVNet-39 training result

Acc@1	FPS	Npu_nums	Epochs	Initial LR	AMP_Type
-	892	1	1	0.0125	O2
77.053	4445	8	90	0.1	O2

Statement

For details about the public address of the code in this repository, you can get from the file public_address_statement.md