vllm: error: unrecognized arguments: --no-enable-chunked-prefill
The error “vllm: error: unrecognized arguments: --no-enable-chunked-prefill” means that the CLI argument --no-enable-chunked-prefill is not recognized by your vLLM version or the specific command you are running. This can happen if the argument is not supported in your vLLM version, or if it is not available for the subcommand you are using. You can check available arguments for your command with vllm serve --help or by consulting the official CLI documentation.
Additionally, in vLLM V1, chunked prefill is always enabled by default and cannot be disabled via CLI arguments; the --no-enable-chunked-prefill flag is ignored or unsupported in this architecture. If you need to disable chunked prefill, you must use the V0 engine by setting the environment variable VLLM_USE_V1=0 before launching vLLM, as described in this issue and here.
Would you like more detail on how to switch between V0 and V1 engines or on chunked prefill behavior?
Sources: