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 7, 06:30 UTC
[MISC] fixed tool_parser mypy errors (#35640)
Signed-off-by: Taneem Ibrahim <taneem.ibrahim@gmail.com> Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com> Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Taneem Ibrahim committed
Mar 4, 2026 at 06:23 UTC
1aaec59d79b8b61c515027320e276d617c4df746
4 files changed
+9
-15
tools/pre_commit/mypy.py
-1
@@ -42,7 +42,6 @@ EXCLUDE = [
42
"vllm/benchmarks",
43
"vllm/config",
44
"vllm/reasoning",
45
- "vllm/tool_parser",
45
]
46
47
vllm/tool_parsers/functiongemma_tool_parser.py
+1
-1
@@ -72,7 +72,7 @@ class FunctionGemmaToolParser(ToolParser):
72
73
def _parse_arguments(self, args_str: str) -> dict:
74
"""Parse FunctionGemma argument string into a dictionary."""
75
- arguments = {}
75
+ arguments: dict = {}
76
if not args_str:
77
return arguments
78
vllm/tool_parsers/glm4_moe_tool_parser.py
+7
-8
@@ -355,12 +355,9 @@ class Glm4MoeModelToolParser(ToolParser):
355
self._buffer = self._buffer[val_end + len(self.arg_val_end) :]
356
self._pending_key = None
357
358
- frag = self._append_arg_fragment(
359
- key=key,
360
- raw_val=raw_val,
361
- )
362
- if frag:
363
- return self._emit_tool_args_delta(frag)
358
+ frag_or_none = self._append_arg_fragment(key=key, raw_val=raw_val)
359
+ if frag_or_none:
360
+ return self._emit_tool_args_delta(frag_or_none)
361
continue
362
363
# Parse next arg or close
@@ -368,7 +365,7 @@ class Glm4MoeModelToolParser(ToolParser):
365
key_pos = self._buffer.find(self.arg_key_start)
366
if end_pos != -1 and (key_pos == -1 or end_pos < key_pos):
367
self._buffer = self._buffer[end_pos + len(self.tool_call_end_token) :]
371
- frag = self._close_args_if_needed()
368
+ frag_or_none = self._close_args_if_needed()
369
# Finalize prev_tool_call_arr with complete parsed arguments
370
if self._current_tool_name:
371
try:
@@ -387,7 +384,9 @@ class Glm4MoeModelToolParser(ToolParser):
384
e,
385
)
386
self._finish_tool_call()
390
- return self._emit_tool_args_delta(frag) if frag else None
387
+ return (
388
+ self._emit_tool_args_delta(frag_or_none) if frag_or_none else None
389
+ )
390
391
if key_pos == -1:
392
return None
vllm/tool_parsers/step3p5_tool_parser.py
+1
-5
@@ -23,10 +23,7 @@ from vllm.entrypoints.openai.engine.protocol import (
23
)
24
from vllm.logger import init_logger
25
from vllm.tokenizers import TokenizerLike
26
-from vllm.tool_parsers.abstract_tool_parser import (
27
- ToolParser,
28
- ToolParserManager,
29
-)
26
+from vllm.tool_parsers.abstract_tool_parser import ToolParser
27
28
logger = init_logger(__name__)
29
@@ -1367,7 +1364,6 @@ class StreamingXMLToolCallParser:
1364
self.deferred_param_raw_value = ""
1365
1366
1370
-@ToolParserManager.register_module("step3p5")
1367
class Step3p5ToolParser(ToolParser):
1368
def __init__(self, tokenizer: TokenizerLike):
1369
super().__init__(tokenizer)