[CI/Build] upgrade compressed-tensors to 0.12.2 to address LGPLv3 #26501
Conversation
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request upgrades the compressed-tensors dependency from version 0.11.0 to 0.12.2 to address a licensing concern. While this is a straightforward version bump in requirements/common.txt, it's important to ensure that there are no regressions or latent bugs in the integration. My review of the related code has uncovered two critical bugs in the compressed-tensors quantization logic that could lead to runtime errors or incorrect behavior. I've detailed these issues in a comment on the requirements file change. It is highly recommended to fix these issues as part of this PR to improve the robustness of the quantization functionality.
csy1204
changed the title
csy1204
requested review from
mgoin,
robertgshaw2-redhat,
tlrmchlsmth and
yewentao256
as code owners
…ency issue Signed-off-by: Sangyeon Cho <josang1204@gmail.com>
Signed-off-by: Sangyeon Cho <josang1204@gmail.com>
Signed-off-by: Sangyeon Cho <josang1204@gmail.com>
There was a problem hiding this comment.
Code Review
This pull request upgrades the compressed-tensors library from version 0.11.0 to 0.12.2, primarily to address a licensing concern. The associated code changes are minimal and correctly adapt the codebase to the new version of the dependency. The modifications in compressed_tensors.py and compressed_tensors_moe.py adjust enum comparisons, which is a common requirement during library upgrades and also fixes a likely pre-existing bug. The changes are sound and the pull request is ready for merging.
| setuptools>=77.0.3,<80; python_version > '3.11' # Setuptools is used by triton, we need to ensure a modern version is installed for 3.12+ so that it does not try to import distutils, which was removed in 3.12 | ||
| einops # Required for Qwen2-VL. | ||
| compressed-tensors == 0.11.0 # required for compressed-tensors | ||
| compressed-tensors == 0.12.2 # required for compressed-tensors |
There was a problem hiding this comment.
This version fixes the issue that occurred in the previous PR
- PR: bumped compressed-tensors version up to 0.12.1 #26173, #ci)
- Fixed PR [Transform] Fix accelerate import to keep it as optional dependency compressed-tensors#480
[2025-10-03T16:37:27Z] ERROR entrypoints/openai/test_response_api_with_harmony.py::test_basic[openai/gpt-oss-20b] - ImportError: cannot import name 'has_offloaded_params' from 'accelerate.utils'
|
@mgoin All CI checks are green. When you get a chance, could you kindly review this PR? Thank you! |
…lm-project#26501) Signed-off-by: Sangyeon Cho <josang1204@gmail.com> Signed-off-by: 1994 <1994@users.noreply.github.com>
…lm-project#26501) Signed-off-by: Sangyeon Cho <josang1204@gmail.com> Signed-off-by: Dhruvil Bhatt <bhattdbh@amazon.com>
…lm-project#26501) Signed-off-by: Sangyeon Cho <josang1204@gmail.com> Signed-off-by: bbartels <benjamin@bartels.dev>
…lm-project#26501) Signed-off-by: Sangyeon Cho <josang1204@gmail.com>
…lm-project#26501) Signed-off-by: Sangyeon Cho <josang1204@gmail.com>
…lm-project#26501) Signed-off-by: Sangyeon Cho <josang1204@gmail.com> Signed-off-by: xuebwang-amd <xuebwang@amd.com>
…lm-project#26501) Signed-off-by: Sangyeon Cho <josang1204@gmail.com> Signed-off-by: xuebwang-amd <xuebwang@amd.com>
…lm-project#26501) Signed-off-by: Sangyeon Cho <josang1204@gmail.com> Signed-off-by: 0xrushi <6279035+0xrushi@users.noreply.github.com>
…lm-project#26501) Signed-off-by: Sangyeon Cho <josang1204@gmail.com> Signed-off-by: 0xrushi <6279035+0xrushi@users.noreply.github.com>
…lm-project#26501) Signed-off-by: Sangyeon Cho <josang1204@gmail.com>
…lm-project#26501) Signed-off-by: Sangyeon Cho <josang1204@gmail.com>
…lm-project#26501) Signed-off-by: Sangyeon Cho <josang1204@gmail.com>
2 participants
Purpose
resolve #26500, vllm-project/compressed-tensors#468
Test Plan
Test Result
tests/quantization/test_compressed_tensors.py result
Essential Elements of an Effective PR Description Checklist
supported_models.mdandexamplesfor a new model.