Deprecated mask fill mask type can causes pages and pages of repeated messages

[hughperkins]

🐛 Bug

When I install DrQA, https://github.com/facebookresearch/DrQA , with pytorch 1.4, the mask being provided is legacy, ie torch.uint8. pytorch correctly points out that this is legacy, and should be migrated to dtype torch.bool. However, rather than printing a single warning, it prints pages and pages of warnings, presumably one for each time the masked fill is called. I feel taht the warning should only be printed once. Example partial output:

/tmp/pip-req-build-o37zi_pa/aten/src/ATen/native/LegacyDefinitions.cpp:19: UserWarning: masked_fill_ received a mask with dtype torch.uint8, this behavior is now deprecated,please use a mask with dtype torch.bool instead.
/tmp/pip-req-build-o37zi_pa/aten/src/ATen/native/LegacyDefinitions.cpp:19: UserWarning: masked_fill_ received a mask with dtype torch.uint8, this behavior is now deprecated,please use a mask with dtype torch.bool instead.
/tmp/pip-req-build-o37zi_pa/aten/src/ATen/native/LegacyDefinitions.cpp:19: UserWarning: masked_fill_ received a mask with dtype torch.uint8, this behavior is now deprecated,please use a mask with dtype torch.bool instead.
/tmp/pip-req-build-o37zi_pa/aten/src/ATen/native/LegacyDefinitions.cpp:19: UserWarning: masked_fill_ received a mask with dtype torch.uint8, this behavior is now deprecated,please use a mask with dtype torch.bool instead.
/tmp/pip-req-build-o37zi_pa/aten/src/ATen/native/LegacyDefinitions.cpp:19: UserWarning: masked_fill_ received a mask with dtype torch.uint8, this behavior is now deprecated,please use a mask with dtype torch.bool instead.
/tmp/pip-req-build-o37zi_pa/aten/src/ATen/native/LegacyDefinitions.cpp:19: UserWarning: masked_fill_ received a mask with dtype torch.uint8, this behavior is now deprecated,please use a mask with dtype torch.bool instead.
/tmp/pip-req-build-o37zi_pa/aten/src/ATen/native/LegacyDefinitions.cpp:19: UserWarning: masked_fill_ received a mask with dtype torch.uint8, this behavior is now deprecated,please use a mask with dtype torch.bool instead.
/tmp/pip-req-build-o37zi_pa/aten/src/ATen/native/LegacyDefinitions.cpp:19: UserWarning: masked_fill_ received a mask with dtype torch.uint8, this behavior is now deprecated,please use a mask with dtype torch.bool instead.

I have logged the masked fill dtype issue at facebookresearch/DrQA#249 . However, I feel taht DrQA cannot simply upgrade this, because will likely break backwards compatibility with earlier torch versions.

I feel that the correct fix is that deprecation warnings should be printed only once per python process.

To Reproduce

Steps to reproduce the behavior:

1. install DrQA, https://github.com/facebookresearch/DrQA (which is a facebook repo actually...)
2. run, using python scripts/pipeline/interactive.py --tokenizer regexp
3. type process('what is the capital of france?')
4. watch the spammy deprecation messages...

Expected behavior

Print the deprecation message once, and then stop.

Environment

Please copy and paste the output from our
environment collection script
(or fill out the checklist below manually).

You can get the script and run it with:

wget https://raw.githubusercontent.com/pytorch/pytorch/master/torch/utils/collect_env.py
# For security purposes, please check the contents of collect_env.py before running it.
python collect_env.py

• PyTorch Version (e.g., 1.0): 1.4
• OS (e.g., Linux): ubuntu 18.04
• How you installed PyTorch (conda, pip, source): conda
• Build command you used (if compiling from source): n/a
• Python version: 3.6
• CUDA/cuDNN version: 10.2
• GPU models and configuration: cpu only
• Any other relevant information:

Additional context

[mruberry]

Thanks for reporting this issue, @hughperkins! I'll take a look.