mlflow cannot serve for vllm and transformers version conflicts

[jeffreyji666]

System Info / 系統信息

试图serve glm ocr by mlflow 3.x, 遇到以下异常:

pip is looking at multiple versions of vllm to determine which version is compatible with other requirements. This could take a while.\nCollecting vllm>=0.17.0 (from -r /local_disk0/user_tmp_data/spark-041cc5c2-f3a5-4ed1-bf60-51/tmp0_agfi8g/model/requirements.txt (line 15))\n Downloading
https://pypi.org/artifactory/api/pypi/pypi-release/packages/packages/4f/e9/59cf9b8939b51e859d2166ac3336b353f52ec4f9ceda34228aae7b386840/vllm-0.18.0-cp38-abi3-manylinux_2_31_x86_64.whl (433.2 MB)\n Downloading https://pypi.org/artifactory/api/pypi/pypi-release/packages/packages/18/28/f85e67b390082481298e27a5c9f1da540d2d5abb1a06a594545cdc320818/vllm-0.17.1-cp38-abi3-manylinux_2_31_x86_64.whl (432.9 MB)\n Downloading https://pypi.org/artifactory/api/pypi/pypi-release/packages/packages/25/4f/972726f9a501f01203b5c4796e1932abbe435fae6d7715a4c3f1aad14a58/vllm-0.17.0-cp38-abi3-manylinux_2_31_x86_64.whl (432.9 MB)\n\x1b[31mERROR: Cannot install -r /local_disk0/user_tmp_data/spark-041cc5c2-f3a5-4ed1-bf60-51/tmp0_agfi8g/model/requirements.txt (line 15) and transformers>=5.3.0 because these package versions have conflicting dependencies.\x1b[0m\x1b[31m\n\x1b[0m\nThe conflict is caused by:\n The user requested transformers>=5.3.0\n vllm 0.18.1 depends on transformers=4.56.0\n The user requested transformers>=5.3.0\n vllm 0.18.0 depends on transformers=4.56.0\n The user requested transformers>=5.3.0\n vllm 0.17.1 depends on transformers=4.56.0\n The user requested transformers>=5.3.0\n vllm 0.17.0 depends on transformers=4.56.0\n\nTo fix this you could try to:\n1. loosen the range of package versions you've specified\n2. remove package versions to allow pip to attempt to solve the dependency conflict\n\n\x1b[31mERROR: ResolutionImpossible: for help visit https://pip.pypa.io/en/latest/topics/dependency-resolution/#dealing-with-dependency-conflicts\x1b[0m\x1b[31m\n\x1b[0m

当通过fastapi serve glm ocr的时候，虽然也报transformers版本不兼容，但仍然安装成功.

Who can help? / 谁可以帮助到您？

No response

Information / 问题信息

• The official example scripts / 官方的示例脚本
• My own modified scripts / 我自己修改的脚本和任务

Reproduction / 复现过程

1. requirements.txt里指定版本：
2. 通过python process 启动 vllm
3. 通过mlflow serve glm, predict方法中调用Pipeline进行OCR识别.

Expected behavior / 期待表现

修复版本冲突，mlflow可以正常serve glm ocr.