This repository has been archived by the owner on Nov 3, 2023. It is now read-only.
Parallelize detensorizing context over batch #3729
Merged
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
stephenroller
approved these changes
Jun 16, 2021
parlai/core/torch_generator_agent.py
Outdated
| """ | ||
| if self.beam_context_block_ngram <= 0: | ||
| # We aren't context blocking, return empty tensor of the correct size | ||
| return torch.LongTensor([[]] * batch.batchsize) |
There was a problem hiding this comment.
Suggested change
| return torch.LongTensor([[]] * batch.batchsize) | |
| return torch.zeros(batch.batchsize, 0) |
There was a problem hiding this comment.
Yeah, that's cleaner - just added that, but with dtype=torch.long
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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Outdated suggestions cannot be applied.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.
Patch description
Currently, when generating, we convert the context Tensor to a list one example at a time, which is expensive; this PR parallelizes this operation over the entire batch.
Performance on 1000 generations (
-mf zoo:blender/blender_90M/model -t blended_skill_talk -ne 1000), on an otherwise unoccupied devfair:Testing steps
CI checks (which seem to cover context blocking)