Zaida Zhou

> Hi @zhouzaida, Is the deadlock can also occur with single gpu training? I'm wonder about the root cause of inserting time.sleep(2) in mmcv. Hi @JihwanEom , in most cases...

Hi, we have not supported loading datasets from Mino but you can refer to [this](https://github.com/open-mmlab/mmcv/blob/6a03918f55c1cc860bcb4d64a0e868ca13cf08ba/mmcv/fileio/file_client.py#L86) to customize a new backend.

Hi @gasvn , thanks for your contributions. Please add docstring and type hints for your added files.

> Is there anything I can do to make some progress? Thanks~ Hi @gasvn , sorry for my late reply. I'm thinking about where to put the implementation of this...

Hi @sleep3r ,thanks for your contribution. What is your PyTorch version?

Thank you very much for the description. BTW, is it convenient for you to test that torch

> @zhouzaida I've just tested. So, seems like this behaviour starts from torch==1.9.0. > > 1.8.0 and 1.8.1 are fine Got it.

Hi @sleep3r , unit tests failed. https://app.circleci.com/pipelines/github/open-mmlab/mmcv/582/workflows/fe9adf11-579c-45c9-8802-4a15f74bb9d9/jobs/1207?invite=true#step-110-124

Hi @filaPro , thanks for your contributions. Could you add some benchmark results in the PR description? You can refer to https://github.com/open-mmlab/mmcv/pull/1718#issuecomment-1061531123.

Please also update https://github.com/open-mmlab/mmcv/blob/master/docs/zh_cn/community/contributing.md