fix: set n_ctx=512 for TinyStories models #1162
Open
Conversation
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
TinyStories models were trained with sequence length 512, but HuggingFace config claims n_ctx=2048. This causes performance degradation for sequences >512 tokens. Added warning to alert users of this limitation. Note: We cannot change n_ctx in the config because the pretrained weights have positional embeddings for 2048 positions. Changing n_ctx would break weight loading. Fixes TransformerLensOrg#492
puranikyashaswin
force-pushed
the
fix/tinystories-n_ctx-492
branch
from
aeca4d1 to
4d74ac0
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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.
Description
TinyStories models were trained with sequence length 512, but HuggingFace config incorrectly claims
n_ctx=2048. This causes severe performance degradation for sequences >512 tokens.This fix adds a post-processing override to correct the
n_ctxvalue when loading any TinyStories model.Fixes #492
Type of change
Checklist: