Add bias to gptj #1363
Merged
Add bias to gptj #1363
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
Transformer4.45 has the Dynamic cache updates removing self.bias causing the failure. Until we investigate further and update the code based on the new transformer, we are putting bias back to GaudiGPTJAttention.init() huggingface/transformers#31421
regisss
approved these changes
Sep 26, 2024
Contributor
Author
|
@atakaha Could you check on this PR? Transformer4.45 has the Dynamic cache updates (huggingface/transformers#31421) which removed self.bias, and that caused an error on our side. I tried to update the model file based on their new logic but found that update_sincos_cache() also using self.bias as well. For now I put the self.bias back to our model file so the test goes through. Could you please review the transformer PR31421 and then see if you can update the gptj model file accordingly? |
Contributor
|
looks good to me. |
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Outdated suggestions cannot be applied.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.
What does this PR do?
Transformer4.45 has the Dynamic cache updates(huggingface/transformers#31421) removing
self.bias causing the failure. Until we investigate further and update the code based on the new transformer, we are putting bias back to GaudiGPTJAttention.init()
Fixes # (issue)
GAUDI2_CI=1 RUN_SLOW=1 python3.10 -m pytest tests/test_text_generation_example.py -k token0-EleutherAI/gpt-j-6b-1