Skip to content
New issue

Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.

#

By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? # to your account

Jump to bottom

[Model] Enable quantization support for transformers backend #12960

Merged
merged 11 commits into from
Feb 18, 2025
Merged

[Model] Enable quantization support for transformers backend #12960

merged 11 commits into from
Feb 18, 2025

Conversation

Isotr0py
Copy link
Collaborator

@Isotr0py Isotr0py commented Feb 8, 2025 •
edited
Loading

  • Add BNB support for transformers backend
  • Update the available quantization in transformers backend docs.

@Isotr0py
add BNB support
92d25f4 
Signed-off-by: Isotr0py <2037008807@qq.com>
@Isotr0py
update annotation
5958fda 
Signed-off-by: Isotr0py <2037008807@qq.com>
@github-actions GitHub Actions
Copy link

github-actions bot commented Feb 8, 2025

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

@mergify mergify bot added the documentation Improvements or additions to documentation label Feb 13, 2025
@Isotr0py
add kv_cache quantization
f07ddb5 
Signed-off-by: isotr0py <2037008807@qq.com>
@Isotr0py
Merge branch 'vllm-project:main' into transformers-bnb
9f05d9a
@Isotr0py
use SupportsQuant interface
3a10be7 
Signed-off-by: Isotr0py <2037008807@qq.com>
@Isotr0py
add bnb test
fba50fb 
Signed-off-by: Isotr0py <2037008807@qq.com>
@Isotr0py
cleanup
acaa322 
Signed-off-by: Isotr0py <2037008807@qq.com>
@Isotr0py Isotr0py marked this pull request as ready for review February 17, 2025 05:37
@Isotr0py Isotr0py requested a review from jeejeelee February 17, 2025 05:37
DarkLight1337
DarkLight1337 approved these changes Feb 17, 2025
View reviewed changes
Copy link
Member

@DarkLight1337 DarkLight1337 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nice, thanks for implementing this!

@DarkLight1337 DarkLight1337 enabled auto-merge (squash) February 17, 2025 05:43
@github-actions github-actions bot added the ready ONLY add when PR is ready to merge/full CI is needed label Feb 17, 2025
hmellor
hmellor reviewed Feb 17, 2025
View reviewed changes
@Isotr0py Isotr0py disabled auto-merge February 17, 2025 08:44
@Isotr0py
address comments
3fd7a2f 
Signed-off-by: Isotr0py <2037008807@qq.com>
@Isotr0py Isotr0py enabled auto-merge (squash) February 17, 2025 09:00
@hmellor hmellor mentioned this pull request Feb 17, 2025
Closed
SunMarc
SunMarc approved these changes Feb 17, 2025
View reviewed changes
Copy link

@SunMarc SunMarc left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks a lot for your work !

tests/models/test_transformers.py
Comment on lines +83 to +93
@pytest.mark.parametrize("model, quantization_kwargs", [
(
"meta-llama/Llama-3.2-1B-Instruct",
{
"quantization": "bitsandbytes",
"load_format": "bitsandbytes",
},
),
])
@pytest.mark.parametrize("max_tokens", [32])
@pytest.mark.parametrize("num_logprobs", [5])
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Would be nice to add more tests in a follow-up PR to cover all important quant scheme as you told me that it should also work for awq, gptq and more + kvcache quant

@simon-mo simon-mo merged commit 67ef8f6 into vllm-project:main Feb 18, 2025
43 of 45 checks passed
panf2333 pushed a commit to yottalabsai/vllm that referenced this pull request Feb 18, 2025
@Isotr0py @panf2333
[Model] Enable quantization support for transformers backend (vllm-…
ae5c017 
…project#12960)
xjpang pushed a commit to xjpang/vllm that referenced this pull request Feb 20, 2025
@Isotr0py
[Model] Enable quantization support for transformers backend (vllm-…
e9c9abe 
…project#12960)
kerthcet pushed a commit to kerthcet/vllm that referenced this pull request Feb 21, 2025
@Isotr0py @kerthcet
[Model] Enable quantization support for transformers backend (vllm-…
2fd84a0 
…project#12960)
@Isotr0py Isotr0py deleted the transformers-bnb branch February 22, 2025 03:43
Akshat-Tripathi pushed a commit to krai/vllm that referenced this pull request Mar 3, 2025
@Isotr0py @Akshat-Tripathi
[Model] Enable quantization support for transformers backend (vllm-…
d32dd01 
…project#12960)
lk-chen pushed a commit to lk-chen/vllm that referenced this pull request Mar 5, 2025
@Isotr0py @lk-chen
[Model] Enable quantization support for transformers backend (vllm-…
07df848 
…project#12960)

Signed-off-by: Linkun Chen <github@lkchen.net>
Said-Akbar pushed a commit to Said-Akbar/vllm-rocm that referenced this pull request Mar 7, 2025
@Isotr0py @Said-Akbar
[Model] Enable quantization support for transformers backend (vllm-…
783ea95 
…project#12960)

Signed-off-by: saeediy <saidakbarp@gmail.com>
# for free to join this conversation on GitHub. Already have an account? # to comment
Reviewers

@jeejeelee jeejeelee jeejeelee approved these changes

@hmellor hmellor hmellor approved these changes

@DarkLight1337 DarkLight1337 DarkLight1337 approved these changes

@SunMarc SunMarc SunMarc approved these changes

@ywang96 ywang96 Awaiting requested review from ywang96 ywang96 is a code owner

Assignees
No one assigned
Labels
documentation Improvements or additions to documentation ready ONLY add when PR is ready to merge/full CI is needed
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
6 participants
@Isotr0py @jeejeelee @hmellor @DarkLight1337 @SunMarc @simon-mo