16-bit Support and Dynamic Loss Scaling#360
Open
jasonkena wants to merge 10 commits intodbolya:masterfrom
Open
16-bit Support and Dynamic Loss Scaling#360jasonkena wants to merge 10 commits intodbolya:masterfrom
jasonkena wants to merge 10 commits intodbolya:masterfrom
Conversation
Contributor
|
Thanks @jasonkena ! The issue description is interesting, I'd especially welcome the fp16 support! But it is quite impossible to review with all the (false possitive) code changes due to your editor re-formatting. Can you avoid that and submit a PR with only your changes? (no unnecessary whitespace) |
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This utilizes Apex's AMP, to enable 16-bit computation, increasing performance, and lowering GPU memory consumption (it saved me 1 GB with batch size 4). Unfortunately, it doesn't work with
torch.jit.The dynamic loss scaling will potentially fix #359, #340, #318, #316, #222, #186, and #56.
I also patched
dcn_v2.pyto support it, so that YOLACT++ will work.I am very sorry for the awful commit history, it was because the Black code-formatter literally modified all of the code.
Cheers