vllm-project/vllm · error · ValueError

cumem allocator is not supported on current platform.

Error message

cumem allocator is not supported on current platform.

What it means

Error "cumem allocator is not supported on current platform." thrown in vllm-project/vllm.

Source

Thrown at vllm/config/model.py:606

            self.tokenizer_revision = resolve_revision(
                self.tokenizer,
                self.tokenizer_revision,
                self.hf_token,
            )

        if self.enable_sleep_mode:
            if not current_platform.is_sleep_mode_available():
                raise ValueError("Sleep mode is not supported on current platform.")
            if current_platform.is_cuda_alike() and not self.enable_cumem_allocator:
                logger.info_once(
                    "Enabling cumem allocator because sleep mode requires it."
                )
                self.enable_cumem_allocator = True
        if (
            self.enable_cumem_allocator
            and not current_platform.is_cumem_allocator_available()
        ):
            raise ValueError("cumem allocator is not supported on current platform.")

        hf_config = get_config(
            self.hf_config_path or self.model,
            self.trust_remote_code,
            self.revision,
            self.code_revision,
            self.config_format,
            hf_overrides_kw=hf_overrides_kw,
            hf_overrides_fn=hf_overrides_fn,
            token=self.hf_token,
        )
        self.hf_config = hf_config
        if dict_overrides:
            self._apply_dict_overrides(hf_config, dict_overrides)
        self.hf_text_config = get_hf_text_config(self.hf_config)
        self.model_arch_config = self.get_model_arch_config()
        self.attention_chunk_size = getattr(
            self.hf_text_config, "attention_chunk_size", None

View on GitHub (pinned to c794754062)

Solutions

1. Do not enable the cumem allocator on the current platform.
2. Run on a platform that supports cuMem (CUDA) if the allocator is required.

When it happens

Trigger: Raised at vllm/config/model.py:606 when validation fails: cumem allocator is not supported on current platform.. 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:606 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/6f8154e8df1f5b73. Report an issue: GitHub.