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 3, 21:30 UTC
[Bugifx] [Qwen3CoderTool] Restore supports_required_and_named for required tool_choice (#42292)
Signed-off-by: chaunceyjiang <chaunceyjiang@gmail.com>
Chauncey committed
May 12, 2026 at 10:09 UTC
920bf3ec8467e5c1007579ec96457d2097238763
1 file changed
+2
-1
vllm/tool_parsers/qwen3coder_tool_parser.py
+2
-1
@@ -19,6 +19,7 @@ from vllm.entrypoints.openai.engine.protocol import (
19
FunctionCall,
20
ToolCall,
21
)
22
+from vllm.envs import VLLM_ENFORCE_STRICT_TOOL_CALLING
23
from vllm.logger import init_logger
24
from vllm.tokenizers import TokenizerLike
25
from vllm.tool_parsers.abstract_tool_parser import (
@@ -35,7 +36,7 @@ logger = init_logger(__name__)
36
37
38
class Qwen3CoderToolParser(ToolParser):
38
- supports_required_and_named: bool = False
39
+ supports_required_and_named: bool = not VLLM_ENFORCE_STRICT_TOOL_CALLING
40
41
def __init__(self, tokenizer: TokenizerLike, tools: list[Tool] | None = None):
42
super().__init__(tokenizer, tools)