vllm-project/vllm · error · ValueError

Cannot use --renderer-num-workers > 1 with the multimodal pr

Error message

Cannot use --renderer-num-workers > 1 with the multimodal processor cache enabled for pooling models. Pooling preprocessing runs on the renderer workers, and the cache is not thread-safe. Please set --renderer-num-workers 1 (the default), or disable the cache with --mm-processor-cache-gb 0.

What it means

Error "Cannot use --renderer-num-workers > 1 with the multimodal processor cache enabled for pooling models. Pooling preprocessing runs on the renderer workers, and the cache is not thread-safe. Please set --renderer-num-workers 1 (the default), or disable the cache with --mm-processor-cache-gb 0." thrown in vllm-project/vllm.

Source

Thrown at vllm/config/model.py:822

            pruning_spec = self.multimodal_config.get_video_pruning_spec()
            supported_pruning = self._model_info.supported_video_pruning_methods
            if (
                pruning_spec is not None
                and supported_pruning
                and pruning_spec[0] not in supported_pruning
            ):
                raise ValueError(
                    f"Video pruning method '{pruning_spec[0]}' is not "
                    f"supported by {self._model_info.architecture} "
                    f"(supported methods: {supported_pruning})."
                )

            if (
                self.renderer_num_workers > 1
                and self.multimodal_config.mm_processor_cache_gb > 0
                and self.runner_type == "pooling"
            ):
                raise ValueError(
                    "Cannot use --renderer-num-workers > 1 with the "
                    "multimodal processor cache enabled for pooling models. "
                    "Pooling preprocessing runs on the renderer workers, and "
                    "the cache is not thread-safe. Please set "
                    "--renderer-num-workers 1 (the default), or "
                    "disable the cache with --mm-processor-cache-gb 0."
                )

            # Rebuild after multimodal_config exists so text-only mm_prefix
            # clearing is applied (and cached for later with_hf_config calls).
            self.model_arch_config = self.get_model_arch_config()

        if self.disable_sliding_window:
            # Set after get_and_verify_max_len to ensure that max_model_len
            # can be correctly capped to sliding window size
            self.hf_text_config.sliding_window = None

        # Avoid running try_verify_and_update_config multiple times

View on GitHub (pinned to c794754062)

Solutions

1. Set --renderer-num-workers 1 (the default) for pooling models with the multimodal processor cache enabled.
2. Or disable the cache with --mm-processor-cache-gb 0.

When it happens

Trigger: Raised at vllm/config/model.py:822 when validation fails: Cannot use --renderer-num-workers > 1 with the multimodal processor cache enabled for pooling models. Typically triggered by an incompatible or incomplete vLLM configuration, an unsupported platform/backend combination, or a runtime resource/dependency that is missing.

Common situations: Commonly encountered at vllm/config/model.py:822 during vLLM startup/config validation or runtime setup when: (1) conflicting CLI flags or config fields are combined, (2) the current platform (CUDA/ROCm/CPU/XPU) or installed optional packages do not support the requested feature, or (3) a required value is absent or out of range. Resolve by correcting the configuration as described in the message, or by selecting a supported alternative.

AI-assisted analysis of vllm-project/vllm@c794754062 (2026-08-14). Data as JSON: /api/errors/bed4728374d61fe2. Report an issue: GitHub.