feat: add yayi2-30b example #208
Closed
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
|
I have added the workaround for the quant as suggested in ml-explore/mlx#328. Now, the example works with 4-bit quantization. Once the PR gets merged, I will upload the 4-bit quantized model. demo.mov |
|
HI @mzbac sorry for the delayed review here. Do you still want to merge this? I think given the non-standard size it wouldn't fit easily in our |
|
I think this should be supported by hf_llm once we fix the quant non-32 dimension. I'm happy to close this one. Meanwhile, if people want to try this model in f16 precision, they should be able to run it via hf_llm. |
|
Sounds good, thank you! |
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Outdated suggestions cannot be applied.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.
Yayi2-30b has a score of over 80 MMLU. I performed some fine-tuning on it using qlora, and from my quick test, it appears to be very promising. so, I created the mlx example for this model in case anyone wants to run it via mlx. However, this model has an unusual k,v layer that causes the quantization to fail. Currently, I haven't found any quantization tools that support this model (except for bnb nf4). It would be great if mlx could provide support for its quantization.
FYI:
https://huggingface.co/wenge-research/yayi2-30b
https://huggingface.co/mzbac/yayi2-30b-guanaco
ml-explore/mlx#328