A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 6, 03:15 UTC
[Frontend][Bugfix] Pass default_chat_template_kwargs to AnthropicServingMessages (#37899)
Signed-off-by: jetxa <jetxzhang@outlook.com>
jetxa committed
Mar 24, 2026 at 13:00 UTC
16a664df240f43c92861e97d5b59a8276e89698b
2 files changed
+3
vllm/entrypoints/anthropic/serving.py
+2
@@ -72,6 +72,7 @@ class AnthropicServingMessages(OpenAIServingChat):
72
tool_parser: str | None = None,
73
enable_prompt_tokens_details: bool = False,
74
enable_force_include_usage: bool = False,
75
+ default_chat_template_kwargs: dict[str, Any] | None = None,
76
):
77
super().__init__(
78
engine_client=engine_client,
@@ -87,6 +88,7 @@ class AnthropicServingMessages(OpenAIServingChat):
88
tool_parser=tool_parser,
89
enable_prompt_tokens_details=enable_prompt_tokens_details,
90
enable_force_include_usage=enable_force_include_usage,
91
+ default_chat_template_kwargs=default_chat_template_kwargs,
92
)
93
self.stop_reason_map = {
94
"stop": "end_turn",
vllm/entrypoints/openai/generate/api_router.py
+1
@@ -150,6 +150,7 @@ async def init_generate_state(
150
reasoning_parser=args.structured_outputs_config.reasoning_parser,
151
enable_prompt_tokens_details=args.enable_prompt_tokens_details,
152
enable_force_include_usage=args.enable_force_include_usage,
153
+ default_chat_template_kwargs=args.default_chat_template_kwargs,
154
)
155
if "generate" in supported_tasks
156
else None