reabse 2026/02/10

Signed-off-by: tzhouam <tzhouam@connect.ust.hk>

docker/Dockerfile.ci

@@ -18,7 +18,7 @@ RUN apt-get update && \
 # Note: --prerelease=allow is required because nightly/commit wheels use
 # pre-release version numbers (e.g. 2.10.0.dev...) which uv ignores by default,
 # causing it to fall back to the older stable release from PyPI.
-RUN uv pip install --system --upgrade --force-reinstall --prerelease=allow vllm --extra-index-url https://wheels.vllm.ai/f97ca671766c5201404e9fc812e35bf2c4e95a01
+RUN uv pip install --system --upgrade --force-reinstall --prerelease=allow vllm --extra-index-url https://wheels.vllm.ai/e94ec597334d9a3e9b0d04bc17152e2747c83d51
 
 # Keep flashinfer runtime packages in version sync.
 # We derive the version from installed flashinfer-python (pulled by vllm)

vllm_omni/config/model.py

@@ -95,6 +95,7 @@ class OmniModelConfig(ModelConfig):
     def __post_init__(
         self,
         # Multimodal config init vars
+        language_model_only: bool,
         limit_mm_per_prompt: dict[str, int | dict[str, int]] | None,
         enable_mm_embeds: bool | None,
         media_io_kwargs: dict[str, dict[str, Any]] | None,
@@ -111,6 +112,7 @@ class OmniModelConfig(ModelConfig):
     ) -> None:
         # Call parent's __post_init__ to handle all standard ModelConfig initialization
         super().__post_init__(
+            language_model_only=language_model_only,
             limit_mm_per_prompt=limit_mm_per_prompt,
             enable_mm_embeds=enable_mm_embeds,
             media_io_kwargs=media_io_kwargs,

vllm_omni/engine/arg_utils.py

@@ -132,6 +132,7 @@ class OmniEngineArgs(EngineArgs):
             skip_tokenizer_init=self.skip_tokenizer_init,
             enable_prompt_embeds=self.enable_prompt_embeds,
             served_model_name=self.served_model_name,
+            language_model_only=self.language_model_only,
             limit_mm_per_prompt=self.limit_mm_per_prompt,
             enable_mm_embeds=self.enable_mm_embeds,
             interleave_mm_strings=self.interleave_mm_strings,
@@ -257,6 +258,7 @@ class AsyncOmniEngineArgs(AsyncEngineArgs):
             skip_tokenizer_init=self.skip_tokenizer_init,
             enable_prompt_embeds=self.enable_prompt_embeds,
             served_model_name=self.served_model_name,
+            language_model_only=self.language_model_only,
             limit_mm_per_prompt=self.limit_mm_per_prompt,
             enable_mm_embeds=self.enable_mm_embeds,
             interleave_mm_strings=self.interleave_mm_strings,