This repository has been archived by the owner on Jun 24, 2024. It is now read-only.
Support for Llama 70-B #409
Merged
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
Expose the sampler configuration structures to allow more flexibility. Add more documentation and description for the sampling functions and structures. Create specific enums for sampler construction and sampling errors. Set n_vocab for the Mirostat 1 sampler in a more reliable way.
This commit removes the pinned version of the `half` crate, allowing consumers to resolve to a version of `half` that is compatible with other dependencies in the project. Signed-off-by: Radu Matei <[email protected]>
LLukas22
reviewed
Aug 18, 2023
| @@ -332,6 +332,7 @@ fn enable_cublas(build: &mut cc::Build, out_dir: &Path) { | |||
| .arg("static") | |||
| .arg("--generate-code=arch=compute_52,code=[compute_52,sm_52]") | |||
| .arg("--generate-code=arch=compute_61,code=[compute_61,sm_61]") | |||
| .arg("--generate-code=arch=compute_75,code=[compute_75,sm_75]") | |||
There was a problem hiding this comment.
Is this needed for newer cards?
There was a problem hiding this comment.
Yes I had issues with my card RTX2070, I guess we should also support 8.x architectures like A100
crates/models/llama/src/lib.rs
Outdated
Comment on lines
491
to
498
| enum LlamaModelVersion { | ||
| Model3b, | ||
| Model7b, | ||
| Model13b, | ||
| Model30b, | ||
| Model65b, | ||
| Model70b, | ||
| } |
There was a problem hiding this comment.
Version hints at the actual llama model version (1 or 2), maybe rename it into LlamaModelType?
|
Looks good, some small nitpicks but if the CI passes it should be good to go 👍 |
|
@LLukas22 Thanks for the review 👍🏼 ! |
|
Thanks for implementing this :D |
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.
Hello,
Solves #402 .
This is a temporary fix for supporting the Llama-2 70B model. I wanted to open a draft PR to get your feedbacks on this implementation for supporting the
n_gqaparams :n_gqaas a optional param inModelParametersLlamaModelVersionenum akin to other e_model union in llama.cppn_head_kvforKandVinstead ofn_headHere is the
llama-2-70B--chat.ggmlv3.q4_0.binmodel loaded on A100 GPU :