Bump llama-cpp-python to 0.2.23 (NVIDIA & CPU-only, no AMD, no Metal)#4924
Merged
Conversation
oobabooga
added a commit
that referenced
this pull request
Dec 14, 2023
|
Awesome, many thanks!! |
|
After this update my token generation speed seems to be about 10x slower on my 3090 running regular LLaMa models. |
Contributor
|
It has a problem offloading KV cache. I forced it on and it's back to normal but I don't think most users can do that. All other models will be 1/2 speed. |
Owner
Author
|
That's a bit of a conundrum because the previous version does not support Mixtral. @Ph0rk0z is it necessary to recompile llama-cpp-python to apply this fix? Can it be monkeypatched? |
Contributor
|
It's not the lib and just the python files. You can edit it under site packages I think. It's not a big patch, it's up in the reddit thread. |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds Mixtral support.
Compiled using GitHub Actions workflows at https://github.com/oobabooga/llama-cpp-python-cuBLAS-wheels
The AMD and Metal workflows are failing, so I only have the NVIDIA and CPU wheels for now.