[Core] Add update_config RPC method

vllm/v1/worker/gpu_model_runner.py

@@ -2,6 +2,7 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 import copy
+import dataclasses
 import gc
 import time
 import weakref
@@ -1692,6 +1693,15 @@ def generate_draft_token_ids(
                 draft_token_ids.append(drafter_output.tolist())
         return draft_token_ids
 
+    def update_config(self, overrides: dict[str, Any]) -> None:
+        for config_name, config_overrides in overrides.items():
+            try:
+                config = getattr(self, config_name)
+            except AttributeError as exc:
+                raise ValueError(f"Unknown config {config_name}") from exc
+            new_config = dataclasses.replace(config, **config_overrides)
+            setattr(self, config_name, new_config)
+
     def load_model(self) -> None:
         logger.info("Starting to load model %s...", self.model_config.model)
         with DeviceMemoryProfiler() as m:  # noqa: SIM117

vllm/v1/worker/gpu_worker.py

@@ -3,7 +3,7 @@
 """A GPU worker class."""
 import gc
 import os
-from typing import TYPE_CHECKING, Optional
+from typing import TYPE_CHECKING, Any, Optional
 
 import torch
 import torch.distributed
@@ -184,6 +184,9 @@ def load_model(self) -> None:
         with context:
             self.model_runner.load_model()
 
+    def update_config(self, overrides: dict[str, Any]) -> None:
+        self.model_runner.update_config(overrides)
+
     @torch.inference_mode()
     def determine_available_memory(self) -> int:
         """Profiles the peak memory usage of the model to determine how much 

vllm/v1/worker/tpu_model_runner.py

@@ -1,9 +1,10 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 import bisect
+import dataclasses
 import gc
 import time
-from typing import TYPE_CHECKING, Optional, cast
+from typing import TYPE_CHECKING, Any, Optional, cast
 from unittest.mock import patch
 
 import numpy as np
@@ -968,6 +969,15 @@ def execute_model(
 
         return model_runner_output
 
+    def update_config(self, overrides: dict[str, Any]) -> None:
+        for config_name, config_overrides in overrides.items():
+            try:
+                config = getattr(self, config_name)
+            except AttributeError as exc:
+                raise ValueError(f"Unknown config {config_name}") from exc
+            new_config = dataclasses.replace(config, **config_overrides)
+            setattr(self, config_name, new_config)
+
     def load_model(self) -> None:
         self.device = self.device_config.device
 

vllm/v1/worker/tpu_worker.py

@@ -2,7 +2,7 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 """A TPU worker class."""
 import os
-from typing import Optional
+from typing import Any, Optional
 
 import torch
 import torch.distributed
@@ -248,6 +248,9 @@ def add_lora(self, lora_request: LoRARequest) -> bool:
     def load_model(self) -> None:
         self.model_runner.load_model()
 
+    def update_config(self, overrides: dict[str, Any]) -> None:
+        self.model_runner.update_config(overrides)
+
     def compile_or_warm_up_model(self) -> None:
         if not self.model_config.enforce_eager:
             self.model_runner.capture_model()