From a5603f21ae7a8dc6f121cba3672123ca34319f43 Mon Sep 17 00:00:00 2001 From: Bugen Zhao Date: Thu, 24 Sep 2026 13:43:24 +0800 Subject: [PATCH] [Rust Frontend] Recognize new frontend-owned serve args as unsupported or no-op (#58330) Co-authored-by: Claude Opus 5.5 Signed-off-by: Bugen Zhao --- rust/src/cmd/src/cli/unsupported.rs | 78 +++++++++++++++++++++++++++++ 1 file changed, 78 insertions(+) diff --git a/rust/src/cmd/src/cli/unsupported.rs b/rust/src/cmd/src/cli/unsupported.rs index 8e514c9534..5088063914 100644 --- a/rust/src/cmd/src/cli/unsupported.rs +++ b/rust/src/cmd/src/cli/unsupported.rs @@ -246,6 +246,13 @@ pub struct EngineUnsupportedArgs { #[arg(long)] pub io_processor_plugin: Option, + /// Number of worker threads in the renderer thread pool. The pool is + /// consumed by the async renderer path to parallelize tokenization, chat + /// template rendering, and multimodal preprocessing across concurrent + /// requests. + #[arg(long)] + pub renderer_num_workers: Option, + /// Path to a dynamically reasoning parser plugin that can be dynamically /// loaded and registered. #[arg(long)] @@ -281,6 +288,16 @@ pub struct EngineUnsupportedArgs { )] pub data_parallel_external_lb: Option, + /// Enable fault tolerance for detailed error recovery, such as scaling + /// down fault DPEngineCore. + #[arg( + long, + visible_alias = "no-enable-fault-tolerance", + default_missing_value = "true", + num_args = 0..=1 + )] + pub enable_fault_tolerance: Option, + /// This feature is work in progress and no prefill optimization takes place /// with this flag enabled currently. #[arg( @@ -308,6 +325,28 @@ pub struct EngineUnsupportedArgs { #[arg(long)] pub mm_processor_cache_type: Option, + /// Hash algorithm to use for multi-modal input caching. Use `"sha256"` or + /// `"sha512"` for FIPS-compliant deployments. + #[arg(long)] + pub mm_hasher_algorithm: Option, + + /// IPC (inter-process communication) method for multimodal tensors. + /// - "direct_rpc": Use msgspec serialization via RPC + /// - "torch_shm": Use torch.multiprocessing shared memory for zero-copy IPC + #[arg(long)] + pub mm_tensor_ipc: Option, + + /// Device the HF multi-modal processor runs the image/video transform on. + /// Convenience for `--mm-processor-kwargs '{"device": ...}'`. + #[arg(long)] + pub mm_processor_device: Option, + + /// Amount of GPU memory (in GiB) sequestered on the engine's device for + /// GPU-side multimodal work in the API-server (frontend) process, such as + /// hardware video decoding. + #[arg(long)] + pub mm_ipc_gpu_memory_gb: Option, + /// Dictionary mapping specific modalities to LoRA model paths. #[arg(long)] pub default_mm_loras: Option, @@ -324,6 +363,17 @@ pub struct EngineUnsupportedArgs { #[arg(long)] pub stream_interval: Option, + /// Maximum number of requests that can be in-flight (waiting or running) + /// at the same time. When the limit is reached, new requests are rejected + /// with HTTP 503 so the client can retry on another instance. + #[arg(long)] + pub max_num_queued_reqs: Option, + + /// Maximum total prompt tokens of requests currently in the prefill phase. + /// When the limit is reached, new requests are rejected with HTTP 503. + #[arg(long)] + pub max_num_queued_tokens: Option, + /// Structured outputs configuration. #[arg(long)] pub structured_outputs_config: Option, @@ -419,6 +469,15 @@ pub struct ServerUnsupportedArgs { #[arg(long)] pub max_log_len: Option, + /// If set to True, include per-request timing metrics in API responses. + #[arg( + long, + visible_alias = "no-enable-per-request-metrics", + default_missing_value = "true", + num_args = 0..=1 + )] + pub enable_per_request_metrics: Option, + /// If set to True, enable tracking server_load_metrics in the app state. #[arg( long, @@ -492,6 +551,15 @@ pub struct ServerUnsupportedArgs { )] pub tokens_only: Option, + /// Controls the `system_fingerprint` field on responses: `full`, `hash`, + /// `custom`, or `none`. + #[arg(long)] + pub fingerprint_mode: Option, + + /// Literal fingerprint string used when `--fingerprint-mode=custom`. + #[arg(long)] + pub fingerprint_value: Option, + /// Log level for uvicorn. #[arg(long)] pub uvicorn_log_level: Option, @@ -562,4 +630,14 @@ pub struct ServerUnsupportedArgs { num_args = 0..=1 )] pub enable_offline_docs: Option, + + /// If set, run pooling score MaxSim on GPU in the API server process. + /// Can significantly improve late-interaction scoring performance. + #[arg( + long, + visible_alias = "no-enable-flash-late-interaction", + default_missing_value = "true", + num_args = 0..=1 + )] + pub enable_flash_late_interaction: Option, }