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, 19:15 UTC
[CI] Fix mypy pre-commit errors on main (#36882)
Signed-off-by: Thomas Parnell <tpa@zurich.ibm.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Thomas Parnell committed
Mar 12, 2026 at 16:22 UTC
abcffbba8c1b8752915fe8ddbb6c77e1eecd18b5
2 files changed
+6
-3
vllm/tokenizers/kimi_audio.py
+5
-2
@@ -4,6 +4,7 @@
4
5
import contextlib
6
import json
7
+from collections.abc import Sequence
8
from pathlib import Path
9
from typing import Any, overload
10
@@ -299,7 +300,9 @@ class KimiAudioTokenizer(TokenizerLike):
300
tokens = self._maybe_truncate(tokens, max_length)
301
return tokens
302
302
- def decode(self, ids: list[int] | int, skip_special_tokens: bool = False) -> str:
303
+ def decode(
304
+ self, ids: Sequence[int] | int, skip_special_tokens: bool = False
305
+ ) -> str:
306
"""Decode token IDs to text, optionally skipping special tokens."""
307
if isinstance(ids, int):
308
ids = [ids]
@@ -321,7 +324,7 @@ class KimiAudioTokenizer(TokenizerLike):
324
return [self._token_to_id.get(token, self._unk_token_id) for token in tokens]
325
326
def convert_ids_to_tokens(
324
- self, ids: list[int], skip_special_tokens: bool = False
327
+ self, ids: Sequence[int], skip_special_tokens: bool = False
328
) -> list[str]:
329
tokens = []
330
for token_id in ids:
vllm/tool_parsers/abstract_tool_parser.py
+1
-1
@@ -68,7 +68,7 @@ class ToolParser:
68
# tool_choice: "Forced Function" or "required" will override
69
# structured output json settings to make tool calling work correctly
70
request.structured_outputs = StructuredOutputsParams(
71
- json=json_schema_from_tool
71
+ json=json_schema_from_tool # type: ignore[call-arg]
72
)
73
request.response_format = None
74
if isinstance(request, ResponsesRequest):