Please refer to the building documentation of MMCV to install MMCV on NPU devices
Here we use 4 NPUs on your computer to train the model with the following command:
bash tools/dist_train.sh configs/deeplabv3/deeplabv3_r50-d8_4xb2-40k_cityscapes-512x1024.py 4Also, you can use only one NPU to train the model with the following command:
python tools/train.py configs/deeplabv3/deeplabv3_r50-d8_4xb2-40k_cityscapes-512x1024.py| Model | mIoU | Config | Download |
|---|---|---|---|
| deeplabv3 | 78.85 | config | log |
| deeplabv3plus | 79.23 | config | log |
| hrnet | 78.1 | config | log |
| fcn | 74.15 | config | log |
| icnet | 69.25 | config | log |
| pspnet | 77.21 | config | log |
| unet | 68.86 | config | log |
| upernet | 77.81 | config | log |
| apcnet | 78.02 | config | log |
| bisenetv1 | 76.04 | config | log |
| bisenetv2 | 72.44 | config | log |
Notes:
- If not specially marked, the results on NPU with amp are the basically same as those on the GPU with FP32.
All above models are provided by Huawei Ascend group.