Mmdetection: UserWarning: indexing with dtype torch.uint8 is now deprecated, please use a dtype torch.bool instead

Created on 25 Jan 2020 · 6Comments · Source: open-mmlab/mmdetection

Hello,
I have installed:
Pytorch 1.4.0
cuda: 10.1

after running:
python3 tools/train.py ./configs/ssd300_coco.py

Training will be started but with this warning, Is there any Idea to remove it?

2020-01-24 18:31:56,041 - INFO - Start running, host: ali@ali-ubt, work_dir: /home/ali/00_2020/mmdetection/work_dirs/ssd300_coco
2020-01-24 18:31:56,041 - INFO - workflow: [('train', 1)], max: 24 epochs
/pytorch/aten/src/ATen/native/IndexingUtils.h:20: UserWarning: indexing with dtype torch.uint8 is now deprecated, please use a dtype torch.bool instead.
/pytorch/aten/src/ATen/native/IndexingUtils.h:20: UserWarning: indexing with dtype torch.uint8 is now deprecated, please use a dtype torch.bool instead.
/pytorch/aten/src/ATen/native/IndexingUtils.h:20: UserWarning: indexing with dtype torch.uint8 is now deprecated, please use a dtype torch.bool instead.
/pytorch/aten/src/ATen/native/IndexingUtils.h:20: UserWarning: indexing with dtype torch.uint8 is now deprecated, please use a dtype torch.bool instead.

Source

Alibhji

👍4

Most helpful comment

You can add this in train.py while training (to ignore the warning):

import warnings
warnings.filterwarnings("ignore", category=UserWarning)

prateek-77 on 2 Feb 2020

👍31

All 6 comments

A first guess is here, for now, you can try to remove the .type(torch.uint8) and also set related variables to be bool somewhere.
Thanks for your kind notification, we will also fix this warning in the future.

ZwwWayne on 26 Jan 2020

I can confirm this issue. I've tried PyTorch versions from 1.4 to 1.2 and still receive the error. However, with PyTorch 1.1 this warning doesn't occur.

CMobley7 on 29 Jan 2020

👍4

When this change is implemented, can it be back-ported into 1.0?

CMobley7 on 31 Jan 2020

This change should be independent, so you can fix it in your environment.

ZwwWayne on 1 Feb 2020

You can add this in train.py while training (to ignore the warning):

import warnings
warnings.filterwarnings("ignore", category=UserWarning)

prateek-77 on 2 Feb 2020

👍31

The below task also produces this error.

root@1bac3eaaf785:/mmdetection# python tools/train.py configs/htc/htc_x101_64x4d_fpn_20e_16gpu.py
2020-03-10 14:49:23,916 - mmdet - INFO - Environment info:
------------------------------------------------------------
TorchVision: 0.4.1a0+d94043a
OpenCV: 4.2.0
MMCV: 0.3.2
MMDetection: 1.1.0+e08a1ee
MMDetection Compiler: GCC 5.4
MMDetection CUDA Compiler: 10.1