Support for safetensor checkpoints #1028
Draft
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Outdated suggestions cannot be applied.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.
Prior to this PR, our checkpointing code only supports loading .pth files but - in support of Mixtral 8x22B - we had need to load a safetensors file (v0.3 of the Instruct checkpoint, published by Mistral). We additionally noted that - as with .pth files - safetensors checkpoints could be split across multiple files. This PR addresses both cases.
Note we also introduce a new required command line parameter: checkpoint-type, which can take the value pth or safetensors.
Finally, a couple of minor fixes:
Set CPU count to 1 to stop some errors we otherwise see
Check that vocab size is correct instead of assuming so and slicing to it