Use more recent version of vllm

[tleyden]

Checklist

• Read CONTRIBUTING.md, and accept the CLA by including the provided snippet. We will not accept PR without this.

PR Description

Before this change, I was getting the error:

root@f6aeecd7dbf8:/workspace/unmute# ./dockerless/start_llm.sh
++ dirname ./dockerless/start_llm.sh
+ cd ./dockerless/..
+ uv tool run vllm@v0.9.0 serve --model=google/gemma-3-1b-it --max-model-len=8192 --dtype=bfloat16 --gpu-memory-utilization=0.3 --port=8091
Installed 149 packages in 638ms
INFO 10-30 20:45:24 [__init__.py:243] Automatically detected platform cuda.
Traceback (most recent call last):
  File "/root/.cache/uv/archive-v0/I7XKAyY3qSMAu9bQ1RA5X/bin/vllm", line 6, in <module>
    from vllm.entrypoints.cli.main import main
  File "/root/.cache/uv/archive-v0/I7XKAyY3qSMAu9bQ1RA5X/lib/python3.12/site-packages/vllm/__init__.py", line 12, in <module>
    from vllm.engine.arg_utils import AsyncEngineArgs, EngineArgs
  File "/root/.cache/uv/archive-v0/I7XKAyY3qSMAu9bQ1RA5X/lib/python3.12/site-packages/vllm/engine/arg_utils.py", line 20, in <module>
    from vllm.config import (BlockSize, CacheConfig, CacheDType, CompilationConfig,
  File "/root/.cache/uv/archive-v0/I7XKAyY3qSMAu9bQ1RA5X/lib/python3.12/site-packages/vllm/config.py", line 38, in <module>
    from vllm.transformers_utils.config import (
  File "/root/.cache/uv/archive-v0/I7XKAyY3qSMAu9bQ1RA5X/lib/python3.12/site-packages/vllm/transformers_utils/config.py", line 31, in <module>
    from vllm.transformers_utils.configs import (ChatGLMConfig, Cohere2Config,
  File "/root/.cache/uv/archive-v0/I7XKAyY3qSMAu9bQ1RA5X/lib/python3.12/site-packages/vllm/transformers_utils/configs/__init__.py", line 26, in <module>
    from vllm.transformers_utils.configs.ovis import OvisConfig
  File "/root/.cache/uv/archive-v0/I7XKAyY3qSMAu9bQ1RA5X/lib/python3.12/site-packages/vllm/transformers_utils/configs/ovis.py", line 75, in <module>
    AutoConfig.register("aimv2", AIMv2Config)
  File "/root/.cache/uv/archive-v0/I7XKAyY3qSMAu9bQ1RA5X/lib/python3.12/site-packages/transformers/models/auto/configuration_auto.py", line 1401, in register
    CONFIG_MAPPING.register(model_type, config, exist_ok=exist_ok)
  File "/root/.cache/uv/archive-v0/I7XKAyY3qSMAu9bQ1RA5X/lib/python3.12/site-packages/transformers/models/auto/configuration_auto.py", line 1081, in register
    raise ValueError(f"'{key}' is already used by a Transformers config, pick another name.")
ValueError: 'aimv2' is already used by a Transformers config, pick another name.

According to vllm-project/vllm-ascend#2046, later versions of vllm fixed the issue.

I verified the fix, now it gets past that error:

(APIServer pid=21797) OSError: You are trying to access a gated repo.
(APIServer pid=21797) Make sure to have access to it at https://huggingface.co/google/gemma-3-1b-it.
(APIServer pid=21797) 401 Client Error. (Request ID: Root=1-6903cf4c-20f4c78f03d4b2ae361ee686;2b90f892-c001-4988-b44e-4375d64090e2)
(APIServer pid=21797)
(APIServer pid=21797) Cannot access gated repo for url https://huggingface.co/google/gemma-3-1b-it/resolve/main/config.json.
(APIServer pid=21797) Access to model google/gemma-3-1b-it is restricted. You must have access to it and be authenticated to access it. Please log in.

[vvolhejn]

Thank you, could you update all occurrences? Try searching for "0.9.1".

[tleyden]

Sure, no problem. I'll update the PR

[tleyden]

Done

[vvolhejn]

Sorry, I think you're still missing start_llm.sh. Happy to merge afterwards!

[tleyden]

I just double checked, and it does look like it was updated in my latest commit: https://github.com/kyutai-labs/unmute/pull/150/files#diff-7087b91bbe9bc682a0e2720f057132e100923f2d63274fe0e42085b130bef6afL5

Can you take another look?

[vvolhejn]

Oh, weird, I must've been looking at an outdated version. Sorry!

(CLA accepted in #149.)