Add pydocstyle to the ruff rules (#52136)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Signed-off-by: Harry Mellor <hej.mellor@gmail.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Harry Mellor
2026-09-16 16:44:29 +00:00
committed by GitHub
co-authored by Claude Opus 5
parent 804e5377aa
commit 8c1557a79c
2002 changed files with 15024 additions and 14269 deletions
@@ -259,7 +259,7 @@ async def _generator(prompt_idx: int, text: str) -> AsyncGenerator[RequestOutput
@pytest.mark.asyncio
@pytest.mark.skip_global_cleanup
async def test_batched_echo_does_not_prepend_user_prompt() -> None:
"""echo must only echo the assistant turn, never the user's prompt.
"""Echo must only echo the assistant turn, never the user's prompt.
With `add_generation_prompt` (the default) the response role is `assistant`
while the last conversation message is the user's, so there is nothing to
@@ -121,7 +121,7 @@ def _build_serving_chat(engine: AsyncLLM) -> OpenAIServingChat:
@pytest.mark.asyncio
async def test_chat_error_non_stream():
"""test finish_reason='error' returns 500 InternalServerError (non-streaming)"""
"""Test finish_reason='error' returns 500 InternalServerError (non-streaming)."""
mock_engine = MagicMock(spec=AsyncLLM)
mock_engine.errored = False
mock_engine.model_config = MockModelConfig()
@@ -246,7 +246,7 @@ async def test_renderer_only_chat_request_skips_mm_cache():
@pytest.mark.asyncio
async def test_chat_error_stream():
"""test finish_reason='error' returns 500 InternalServerError (streaming)"""
"""Test finish_reason='error' returns 500 InternalServerError (streaming)."""
mock_engine = MagicMock(spec=AsyncLLM)
mock_engine.errored = False
mock_engine.model_config = MockModelConfig()
@@ -465,7 +465,8 @@ def test_system_message_warns_on_video(video_content):
def test_json_schema_response_format_missing_schema():
"""When response_format type is 'json_schema' but the json_schema field
is not provided, request construction should raise a validation error
so the API returns 400 instead of 500."""
so the API returns 400 instead of 500.
"""
with pytest.raises(Exception, match="json_schema.*must be provided"):
ChatCompletionRequest(
model=MODEL_NAME,
@@ -519,7 +520,8 @@ def test_structured_outputs_structural_tag_invalid(structural_tag):
@pytest.mark.parametrize("field_name", ["prompt_logprobs", "top_logprobs"])
def test_non_numeric_logprobs_rejected(field_name):
"""A non-numeric logprobs value must be a clean 400 validation error, not a
TypeError from the mode='before' comparison (which surfaces as HTTP 500)."""
TypeError from the mode='before' comparison (which surfaces as HTTP 500).
"""
with pytest.raises(VLLMValidationError, match=f"`{field_name}` must be an integer"):
ChatCompletionRequest(
model=MODEL_NAME,
@@ -81,7 +81,8 @@ async def test_chat_logit_bias_invalid(client):
@pytest.mark.asyncio
async def test_chat_logit_bias_non_integer_key(client):
"""Test that a non-integer logit_bias key is rejected with a clean,
informative error instead of a raw 'invalid literal for int()' message."""
informative error instead of a raw 'invalid literal for int()' message.
"""
with pytest.raises(openai.BadRequestError) as excinfo:
await client.chat.completions.create(
model=MODEL_NAME,
@@ -101,7 +102,8 @@ async def test_chat_logit_bias_non_integer_key(client):
@pytest.mark.asyncio
async def test_chat_logit_bias_non_numeric_value(client):
"""Test that a non-numeric logit_bias value is rejected with a message
that names the specific offending token, not just a generic TypeError."""
that names the specific offending token, not just a generic TypeError.
"""
with pytest.raises(openai.BadRequestError) as excinfo:
await client.chat.completions.create(
model=MODEL_NAME,
@@ -120,7 +122,8 @@ async def test_chat_logit_bias_non_numeric_value(client):
@pytest.mark.asyncio
async def test_chat_logit_bias_multiple_non_integer_keys(client):
"""Test that ALL invalid logit_bias keys are reported together,
not just the first one encountered."""
not just the first one encountered.
"""
with pytest.raises(openai.BadRequestError) as excinfo:
await client.chat.completions.create(
model=MODEL_NAME,
@@ -451,7 +451,7 @@ async def test_inconsistent_tool_choice_and_tools(
async def test_max_tokens_with_tool_choice_required(
client: openai.AsyncOpenAI, tool_choice
):
""" """
"""Requesting a single token with a required tool choice must not crash."""
models = await client.models.list()
model_name: str = models.data[0].id
@@ -52,9 +52,10 @@ def stop_sign_image_url():
@pytest.mark.asyncio
async def test_translategemma_extra_lang_code_fields(client: openai.AsyncOpenAI):
"""en -> es translation through TranslateGemma's bundled chat template's
"""En -> es translation through TranslateGemma's bundled chat template's
``text`` content branch, which depends on ``source_lang_code`` /
``target_lang_code`` extra fields being preserved on the content part."""
``target_lang_code`` extra fields being preserved on the content part.
"""
completion = await client.chat.completions.create(
model=MODEL_NAME,
messages=[
@@ -92,13 +93,14 @@ async def test_translategemma_image_extra_lang_code_fields(
client: openai.AsyncOpenAI,
stop_sign_image_url: str,
):
"""en -> es OCR-translation through TranslateGemma's bundled chat
"""En -> es OCR-translation through TranslateGemma's bundled chat
template's ``image`` content branch. Exercises the multimodal branch of
``_collect_extra_fields`` in ``_parse_chat_message_content_part``: the
``source_lang_code`` / ``target_lang_code`` extras must survive parsing
and end up alongside the ``{"type": "image"}`` placeholder so the
template's image branch (``content["type"] == 'image'``) renders the
right OCR-translation prompt."""
right OCR-translation prompt.
"""
completion = await client.chat.completions.create(
model=MODEL_NAME,
messages=[
@@ -191,7 +191,8 @@ async def test_logprob_token_ids_stream_with_no_top_k(server):
async def test_logprob_token_ids_default_behavior_unchanged(server):
"""Without `logprob_token_ids`, the response carries the natural top-k
most-likely tokens. This guards against the new field accidentally
changing the default-path output."""
changing the default-path output.
"""
async with server.get_async_client() as client:
resp = await client.chat.completions.create(
model=MODEL_NAME,
@@ -1232,7 +1232,6 @@ async def test_serving_chat_mistral_token_ids_prompt_is_validated():
"""Regression test: when the Mistral tokenizer path returns token IDs
directly, we must still apply input length + max_tokens validation.
"""
mock_engine = MagicMock(spec=AsyncLLM)
mock_engine.errored = False
mock_engine.model_config = MockModelConfig(skip_tokenizer_init=True)
@@ -1271,7 +1270,6 @@ async def test_serving_chat_mistral_token_ids_prompt_too_long_is_rejected():
"""Regression test: MistralTokenizer token-id prompts must still enforce
the max context length for the input itself (token_num >= max_model_len).
"""
mock_engine = MagicMock(spec=AsyncLLM)
mock_engine.errored = False
mock_engine.model_config = MockModelConfig(skip_tokenizer_init=True)
@@ -1408,7 +1406,8 @@ async def test_serving_chat_did_set_correct_cache_salt(model_type):
@pytest.mark.asyncio
async def test_serving_chat_data_parallel_rank_extraction():
"""Test that data_parallel_rank is properly extracted from header and
passed to engine."""
passed to engine.
"""
mock_engine = MagicMock(spec=AsyncLLM)
mock_engine.errored = False
mock_engine.model_config = MockModelConfig()
@@ -1512,8 +1511,7 @@ async def _render_chat_prompt_token_ids(
class TestServingChatWithHarmony:
"""
These tests ensure Chat Completion requests are being properly converted into
"""These tests ensure Chat Completion requests are being properly converted into
Harmony messages and Harmony response messages back into Chat Completion responses.
These tests are not exhaustive, but each one was created to cover a specific case
that we got wrong but is now fixed.
@@ -1731,7 +1729,8 @@ class TestServingChatWithHarmony:
@pytest.mark.asyncio
async def test_system_message_without_tools(self, serving_chat, stream):
"""Leading system message produces a developer message with
DeveloperContent (# Instructions header)."""
DeveloperContent (# Instructions header).
"""
messages = [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "Hello"},
@@ -1753,7 +1752,8 @@ class TestServingChatWithHarmony:
@pytest.mark.asyncio
async def test_system_message_with_tools(self, serving_chat, stream, weather_tools):
"""Leading system message is folded into the developer message
alongside tool definitions."""
alongside tool definitions.
"""
messages = [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "What's the weather?"},
@@ -2168,7 +2168,8 @@ class TestServingChatWithHarmony:
@pytest.mark.asyncio
async def test_tool_choice_validation_without_parser():
"""Test that tool_choice='required' or named tool without tool_parser
returns an appropriate error message."""
returns an appropriate error message.
"""
mock_engine = MagicMock(spec=AsyncLLM)
mock_engine.errored = False
mock_engine.model_config = MockModelConfig()
@@ -33,7 +33,6 @@ def _count_reasoning_decode_token_ids_between_markers(
reasoning_end_ids: list[int],
) -> int | None:
"""Count decode tokens in the thinking span (after last start, before first end)."""
if not reasoning_start_ids or not reasoning_end_ids:
raise ValueError("reasoning marker token id lists must be non-empty")
@@ -152,8 +151,8 @@ async def client(request, server, server_with_auto_reasoning_config):
@pytest.mark.parametrize("client", ["default", "auto_config"], indirect=True)
async def test_thinking_token_budget_mixed_requests(client: openai.AsyncOpenAI):
"""Test that mixed requests (some with thinking_token_budget, some without)
complete successfully without errors."""
complete successfully without errors.
"""
response_with_budget = await client.chat.completions.create(
model=MODEL_NAME,
messages=MESSAGES,
@@ -182,7 +181,6 @@ async def test_thinking_token_budget_limits_reasoning(client: openai.AsyncOpenAI
grouped into streamed chunks (a single chunk can carry several tokens under
async scheduling / stream_interval > 1). Counting chunks under-counts.
"""
tokenizer = get_tokenizer(tokenizer_name=MODEL_NAME)
start_ids = list(tokenizer.encode(REASONING_START_STR, add_special_tokens=False))
end_ids = list(tokenizer.encode(REASONING_END_STR, add_special_tokens=False))
@@ -227,7 +225,6 @@ async def test_thinking_token_budget_qwen35_fp8_mtp_concurrent_mixed_budget_and_
Qwen3.5 FP8 + MTP (TP=2) server. Budgeted calls are checked with
``_count_reasoning_decode_token_ids_between_markers`` on full token ids.
"""
_batch_spec: list[tuple[Literal["budget"], int] | tuple[Literal["plain"], None]] = [
("budget", 1),
("budget", 12),