llama : add support for GPT2, Bloom and CodeShell tied word embeddings #12456
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
It appears transformer.wte.weight is in the weight map even though the weights are not there, remove it if output weights are encountered first.
ngxson
approved these changes
Mar 19, 2025
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Outdated suggestions cannot be applied.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.
Also remove weight duplication from said models on conversion.
I've converted and tested the following models, confirming that they do not initially have output weights (except for CodeShell, see below) but rely on word embeddings and output weights being tied together at runtime:
For some reason CodeShell has inverted ties; output weights are provided in the bin/safetensors, but not word embeddings, even though our conversion code seems to imply otherwise.
Added a workaround for
transformer.wte.weightbeing in the CodeShell weight map even though it's not in the tensor file(s), causing a conversion error unless you edit the .index.json file.