Set expected place in child thread for dataloader to avoid costing cuda memory on other card #30338
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Thanks for your contribution! |
heavengate
approved these changes
Jan 13, 2021
Xreki
approved these changes
Jan 13, 2021
zhiqiu
added a commit
to zhiqiu/Paddle
that referenced
this pull request
Jan 13, 2021
…da memory on other card (PaddlePaddle#30338) * set expected place in child thread for dataloader * set device id when set tensor from numpy * revert tensor_py change * add compile guard * fix ci * fix bug
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
PR types
Bug fixes
PR changes
Others
Describe
Set expected place in child thread for dataloader to avoid costing cuda memory on other card
cudaSetDevice()is valid inside host thread, so the child thread in dataloader may have the default device id (0) if not set explicitly, which may cost about 500MB memory on card 0.It is worse when doing multi-card training in dygraph, since each process will cost about 500MB memory on card 0.