[fix] Preserve <eos> token and in-place it after trimming
#401
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
jon-tow
approved these changes
Mar 28, 2023
| @@ -445,7 +445,7 @@ def make_experience(self, num_rollouts: int = 1024, iter_count: int = 0): # noq | |||
| sample_outputs = sample_outputs.cpu() | |||
| values = values.cpu()[:, :-1] | |||
|
|
|||
| ends = start + attention_mask[:, start:].sum(1) | |||
| ends = start + attention_mask[:, start:].sum(1) + 1 | |||
There was a problem hiding this comment.
Nit: Maybe add a comment saying something like "add 1 to account for appended eos token" since the decode call is a few blocks up
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Outdated suggestions cannot be applied.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.
This PR fixes #399 and #373 by keeping
<|endoftext|>token if it was outputted by the model (previously it was treated as if it was one ofstop_sequences) and replacing trimming after-and-includingstop_sequenceswith<|endoftext|>, instead of simply discarding them. This lets PPO'smake_experiencenever filter any empty outputs, or in the worst case, stuck and timeout.https://wandb.ai/sorry/trlx-references/reports/fix-preserve-eos-token-v-main--VmlldzozOTE1MDU2