{"record":{"id":"89204ad81e6258a3","repo":"vllm-project/vllm","slug":"stochastic-rounding-for-mamba-cache-is-only-suppor","errorCode":null,"errorMessage":"Stochastic rounding for Mamba cache is only supported on NVIDIA CUDA platforms. Please do not specify  `--enable-mamba-cache-stochastic-rounding`.","messagePattern":"Stochastic rounding for Mamba cache is only supported on NVIDIA CUDA platforms\\. Please do not specify  `--enable-mamba-cache-stochastic-rounding`\\.","errorType":"validation","errorClass":"ValueError","httpStatus":null,"severity":"error","filePath":"vllm/config/mamba.py","lineNumber":89,"sourceCode":"            valid = \", \".join(valid_algorithms)\n            raise ValueError(\n                f\"Unknown Mamba SSU algorithm: '{self.ssu_algorithm}'. \"\n                f\"Valid options are: {valid}\"\n            )\n        if self.backend != MambaBackendEnum.FLASHINFER:\n            raise ValueError(\n                \"Mamba SSU algorithm selection is only supported with the \"\n                \"FlashInfer backend. Please set `--mamba-backend flashinfer`, \"\n                \"or omit `--mamba-ssu-algorithm`.\"\n            )\n\n    def __post_init__(self):\n        self.validate_ssu_algorithm()\n        if self.enable_stochastic_rounding:\n            from vllm.platforms import current_platform\n\n            if not current_platform.is_cuda():\n                raise ValueError(\n                    \"Stochastic rounding for Mamba cache is only supported \"\n                    \"on NVIDIA CUDA platforms. Please do not specify  \"\n                    \"`--enable-mamba-cache-stochastic-rounding`.\"\n                )\n            if (\n                self.backend == MambaBackendEnum.TRITON\n                and not current_platform.is_device_capability_family(100)\n            ):\n                raise ValueError(\n                    \"Stochastic rounding for Mamba cache with triton backend requires \"\n                    \"compute capability 10.0 (data center Blackwell). The `cvt.rs` \"\n                    \"PTX instruction is not supported on your GPU. Please do not \"\n                    \"specify `--enable-mamba-cache-stochastic-rounding`, \"\n                    \"or set `--mamba-backend flashinfer`.\"\n                )\n","sourceCodeStart":71,"sourceCodeEnd":105,"githubUrl":"https://github.com/vllm-project/vllm/blob/c794754062d49a8fdb63ab3c5215b488b865030c/vllm/config/mamba.py#L71-L105","documentation":"Error \"Stochastic rounding for Mamba cache is only supported on NVIDIA CUDA platforms. Please do not specify  `--enable-mamba-cache-stochastic-rounding`.\" thrown in vllm-project/vllm.","triggerScenarios":"Raised at vllm/config/mamba.py:89 when validation fails: Stochastic rounding for Mamba cache is only supported on NVIDIA CUDA platforms. Typically triggered by an incompatible or incomplete vLLM configuration, an unsupported platform/backend combination, or a runtime resource/dependency that is missing.","commonSituations":"Commonly encountered at vllm/config/mamba.py:89 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.","solutions":["Do not specify --enable-mamba-cache-stochastic-rounding on non-CUDA platforms."],"exampleFix":null,"handlingStrategy":null,"validationCode":null,"typeGuard":null,"tryCatchPattern":null,"preventionTips":[],"tags":[],"backgroundTag":null,"analyzedSha":"c794754062d49a8fdb63ab3c5215b488b865030c","analyzedAt":"2026-08-14T21:17:39.825Z","schemaVersion":2},"datasetVersion":"2026-08-15T17:31:12.345Z"}