Add support for GPTQ-quantized MoE models using MoE Marlin #2557
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
Narsil
reviewed
Sep 24, 2024
| @@ -5,7 +5,7 @@ | |||
| inputs.nixpkgs.follows = "tgi-nix/nixpkgs"; | |||
| }; | |||
| nix-filter.url = "github:numtide/nix-filter"; | |||
| tgi-nix.url = "github:danieldk/tgi-nix"; | |||
| tgi-nix.url = "github:danieldk/tgi-nix/moe-kernels-0.4.0"; | |||
There was a problem hiding this comment.
To be moved back on master no ?
There was a problem hiding this comment.
Yeah. On a branch to avoid accidentally updating the tgi-nix input, which would break MoE on TGI master (since 0.4.0 is an API break).
danieldk
force-pushed
the
feature/moe-marlin
branch
from
525e980 to
06f87c3
Compare
danieldk
force-pushed
the
feature/moe-marlin
branch
2 times, most recently
from
6d543b2 to
83953fa
Compare
This change add support for MoE models that use GPTQ quantization. Currently only models with the following properties are supported: - No `desc_act` with tensor parallelism, unless `group_size=-1`. - No asymmetric quantization. - No AWQ.
danieldk
force-pushed
the
feature/moe-marlin
branch
from
83953fa to
5a08c33
Compare
Narsil
approved these changes
Sep 30, 2024
yuanwu2017
pushed a commit
to yuanwu2017/tgi-gaudi
that referenced
this pull request
Oct 27, 2024
…ce#2557) This change add support for MoE models that use GPTQ quantization. Currently only models with the following properties are supported: - No `desc_act` with tensor parallelism, unless `group_size=-1`. - No asymmetric quantization. - No AWQ.
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Outdated suggestions cannot be applied.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.
What does this PR do?
This change add support for MoE models that use GPTQ quantization.
Currently only models with the following properties are supported:
desc_actwith tensor parallelism, unlessgroup_size=-1.Draft: needs to be rebased on #2537 once that PR is merged.
Before submitting
Pull Request section?
to it if that's the case.
documentation guidelines, and
here are tips on formatting docstrings.
Who can review?
Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.