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 10, 10:15 UTC
[Bugfix] Fix OpenAPI schema test failures (#31921)
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
Andreas Karatzas committed
Jan 9, 2026 at 04:56 UTC
020732800caf3ba1eae62098c4264dac7ef35611
2 files changed
+11
-3
tests/entrypoints/openai/test_openai_schema.py
+5
-1
@@ -90,7 +90,10 @@ def before_generate_case(context: schemathesis.hooks.HookContext, strategy):
90
if (
91
isinstance(content, list)
92
and len(content) > 0
93
- and any(item.get("type") == "file" for item in content)
93
+ and any(
94
+ isinstance(item, dict) and item.get("type") == "file"
95
+ for item in content
96
+ )
97
):
98
return False
99
@@ -139,6 +142,7 @@ def test_openapi_stateless(case: schemathesis.Case):
142
timeout = {
143
# requires a longer timeout
144
("POST", "/v1/chat/completions"): LONG_TIMEOUT_SECONDS,
145
+ ("POST", "/v1/completions"): LONG_TIMEOUT_SECONDS,
146
}.get(key, DEFAULT_TIMEOUT_SECONDS)
147
148
# No need to verify SSL certificate for localhost
vllm/entrypoints/openai/protocol.py
+6
-2
@@ -577,7 +577,9 @@ class ChatCompletionRequest(OpenAIBaseModel):
577
min_tokens: int = 0
578
skip_special_tokens: bool = True
579
spaces_between_special_tokens: bool = True
580
- truncate_prompt_tokens: Annotated[int, Field(ge=-1)] | None = None
580
+ truncate_prompt_tokens: Annotated[int, Field(ge=-1, le=_LONG_INFO.max)] | None = (
581
+ None
582
+ )
583
prompt_logprobs: int | None = None
584
allowed_token_ids: list[int] | None = None
585
bad_words: list[str] = Field(default_factory=list)
@@ -1058,7 +1060,9 @@ class CompletionRequest(OpenAIBaseModel):
1060
min_tokens: int = 0
1061
skip_special_tokens: bool = True
1062
spaces_between_special_tokens: bool = True
1061
- truncate_prompt_tokens: Annotated[int, Field(ge=-1)] | None = None
1063
+ truncate_prompt_tokens: Annotated[int, Field(ge=-1, le=_LONG_INFO.max)] | None = (
1064
+ None
1065
+ )
1066
allowed_token_ids: list[int] | None = None
1067
prompt_logprobs: int | None = None
1068
# --8<-- [end:completion-sampling-params]