youngkingdom/vllm
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[Misc][OpenAI] deprecate max_tokens in favor of new max_completion_tokens field for chat completion endpoint (#9837)

Browse Source
This commit is contained in:
Guillaume Calmettes
2024-10-31 02:15:56 +01:00
committed by GitHub
parent 64384bbcdf
commit abbfb6134d
14 changed files with 140 additions and 118 deletions
Download Patch File Download Diff File Expand all files Collapse all files

32
tests/entrypoints/openai/test_audio.py
View File

@ -68,11 +68,12 @@ async def test_single_chat_session_audio(client: openai.AsyncOpenAI,
}]
# test single completion
chat_completion = await client.chat.completions.create(model=model_name,
messages=messages,
max_tokens=10,
logprobs=True,
top_logprobs=5)
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_completion_tokens=10,
logprobs=True,
top_logprobs=5)
assert len(chat_completion.choices) == 1
choice = chat_completion.choices[0]
@ -91,7 +92,7 @@ async def test_single_chat_session_audio(client: openai.AsyncOpenAI,
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
)
message = chat_completion.choices[0].message
assert message.content is not None and len(message.content) >= 0
@ -123,11 +124,12 @@ async def test_single_chat_session_audio_base64encoded(
}]
# test single completion
chat_completion = await client.chat.completions.create(model=model_name,
messages=messages,
max_tokens=10,
logprobs=True,
top_logprobs=5)
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_completion_tokens=10,
logprobs=True,
top_logprobs=5)
assert len(chat_completion.choices) == 1
choice = chat_completion.choices[0]
@ -146,7 +148,7 @@ async def test_single_chat_session_audio_base64encoded(
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
)
message = chat_completion.choices[0].message
assert message.content is not None and len(message.content) >= 0
@ -178,7 +180,7 @@ async def test_chat_streaming_audio(client: openai.AsyncOpenAI,
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
temperature=0.0,
)
output = chat_completion.choices[0].message.content
@ -188,7 +190,7 @@ async def test_chat_streaming_audio(client: openai.AsyncOpenAI,
stream = await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
temperature=0.0,
stream=True,
)
@ -242,7 +244,7 @@ async def test_multi_audio_input(client: openai.AsyncOpenAI, model_name: str,
await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
temperature=0.0,
)

103
tests/entrypoints/openai/test_chat.py
View File

@ -65,11 +65,12 @@ async def test_no_logprobs_chat(client: openai.AsyncOpenAI, model_name: str):
"content": "what is 1+1?"
}]
chat_completion = await client.chat.completions.create(model=model_name,
messages=messages,
max_tokens=5,
temperature=0.0,
logprobs=False)
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_completion_tokens=5,
temperature=0.0,
logprobs=False)
choice = chat_completion.choices[0]
assert choice.logprobs is None
@ -90,12 +91,13 @@ async def test_zero_logprobs_chat(client: openai.AsyncOpenAI, model_name: str):
"content": "what is 1+1?"
}]
chat_completion = await client.chat.completions.create(model=model_name,
messages=messages,
max_tokens=5,
temperature=0.0,
logprobs=True,
top_logprobs=0)
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_completion_tokens=5,
temperature=0.0,
logprobs=True,
top_logprobs=0)
choice = chat_completion.choices[0]
assert choice.logprobs is not None
@ -117,12 +119,13 @@ async def test_some_logprobs_chat(client: openai.AsyncOpenAI, model_name: str):
"content": "what is 1+1?"
}]
chat_completion = await client.chat.completions.create(model=model_name,
messages=messages,
max_tokens=5,
temperature=0.0,
logprobs=True,
top_logprobs=5)
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_completion_tokens=5,
temperature=0.0,
logprobs=True,
top_logprobs=5)
choice = chat_completion.choices[0]
assert choice.logprobs is not None
@ -149,7 +152,7 @@ async def test_too_many_chat_logprobs(client: openai.AsyncOpenAI,
with pytest.raises((openai.BadRequestError, openai.APIError)):
stream = await client.chat.completions.create(model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
logprobs=True,
top_logprobs=21,
stream=True)
@ -159,16 +162,17 @@ async def test_too_many_chat_logprobs(client: openai.AsyncOpenAI,
with pytest.raises(openai.BadRequestError):
await client.chat.completions.create(model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
logprobs=True,
top_logprobs=30,
stream=False)
# the server should still work afterwards
chat_completion = await client.chat.completions.create(model=model_name,
messages=messages,
max_tokens=10,
stream=False)
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_completion_tokens=10,
stream=False)
message = chat_completion.choices[0].message
assert message.content is not None and len(message.content) >= 0
@ -271,11 +275,12 @@ async def test_single_chat_session(client: openai.AsyncOpenAI,
}]
# test single completion
chat_completion = await client.chat.completions.create(model=model_name,
messages=messages,
max_tokens=10,
logprobs=True,
top_logprobs=5)
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_completion_tokens=10,
logprobs=True,
top_logprobs=5)
assert chat_completion.id is not None
assert len(chat_completion.choices) == 1
@ -294,7 +299,7 @@ async def test_single_chat_session(client: openai.AsyncOpenAI,
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
)
message = chat_completion.choices[0].message
assert message.content is not None and len(message.content) >= 0
@ -319,7 +324,7 @@ async def test_chat_streaming(client: openai.AsyncOpenAI, model_name: str):
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
temperature=0.0,
)
output = chat_completion.choices[0].message.content
@ -329,7 +334,7 @@ async def test_chat_streaming(client: openai.AsyncOpenAI, model_name: str):
stream = await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
temperature=0.0,
stream=True,
)
@ -369,7 +374,7 @@ async def test_chat_completion_stream_options(client: openai.AsyncOpenAI,
stream = await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
temperature=0.0,
stream=True,
stream_options={"include_usage": False})
@ -380,7 +385,7 @@ async def test_chat_completion_stream_options(client: openai.AsyncOpenAI,
# "continuous_usage_stats": False}}
stream = await client.chat.completions.create(model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
temperature=0.0,
stream=True,
stream_options={
@ -409,7 +414,7 @@ async def test_chat_completion_stream_options(client: openai.AsyncOpenAI,
await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
temperature=0.0,
stream=False,
stream_options={"include_usage": None})
@ -419,7 +424,7 @@ async def test_chat_completion_stream_options(client: openai.AsyncOpenAI,
await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
temperature=0.0,
stream=False,
stream_options={"include_usage": True})
@ -429,7 +434,7 @@ async def test_chat_completion_stream_options(client: openai.AsyncOpenAI,
stream = await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
extra_body=dict(min_tokens=10),
temperature=0.0,
stream=True,
@ -476,7 +481,7 @@ async def test_guided_choice_chat(client: openai.AsyncOpenAI,
chat_completion = await client.chat.completions.create(
model=MODEL_NAME,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
extra_body=dict(guided_choice=sample_guided_choice,
guided_decoding_backend=guided_decoding_backend))
choice1 = chat_completion.choices[0].message.content
@ -490,7 +495,7 @@ async def test_guided_choice_chat(client: openai.AsyncOpenAI,
chat_completion = await client.chat.completions.create(
model=MODEL_NAME,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
extra_body=dict(guided_choice=sample_guided_choice,
guided_decoding_backend=guided_decoding_backend))
choice2 = chat_completion.choices[0].message.content
@ -517,7 +522,7 @@ async def test_guided_json_chat(client: openai.AsyncOpenAI,
chat_completion = await client.chat.completions.create(
model=MODEL_NAME,
messages=messages,
max_tokens=1000,
max_completion_tokens=1000,
extra_body=dict(guided_json=sample_json_schema,
guided_decoding_backend=guided_decoding_backend))
message = chat_completion.choices[0].message
@ -535,7 +540,7 @@ async def test_guided_json_chat(client: openai.AsyncOpenAI,
chat_completion = await client.chat.completions.create(
model=MODEL_NAME,
messages=messages,
max_tokens=1000,
max_completion_tokens=1000,
extra_body=dict(guided_json=sample_json_schema,
guided_decoding_backend=guided_decoding_backend))
message = chat_completion.choices[0].message
@ -563,7 +568,7 @@ async def test_guided_regex_chat(client: openai.AsyncOpenAI,
chat_completion = await client.chat.completions.create(
model=MODEL_NAME,
messages=messages,
max_tokens=20,
max_completion_tokens=20,
extra_body=dict(guided_regex=sample_regex,
guided_decoding_backend=guided_decoding_backend))
ip1 = chat_completion.choices[0].message.content
@ -575,7 +580,7 @@ async def test_guided_regex_chat(client: openai.AsyncOpenAI,
chat_completion = await client.chat.completions.create(
model=MODEL_NAME,
messages=messages,
max_tokens=20,
max_completion_tokens=20,
extra_body=dict(guided_regex=sample_regex,
guided_decoding_backend=guided_decoding_backend))
ip2 = chat_completion.choices[0].message.content
@ -623,7 +628,7 @@ async def test_guided_choice_chat_logprobs(client: openai.AsyncOpenAI,
chat_completion = await client.chat.completions.create(
model=MODEL_NAME,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
logprobs=True,
top_logprobs=5,
extra_body=dict(guided_choice=sample_guided_choice,
@ -660,7 +665,7 @@ async def test_named_tool_use(client: openai.AsyncOpenAI,
chat_completion = await client.chat.completions.create(
model=MODEL_NAME,
messages=messages,
max_tokens=1000,
max_completion_tokens=1000,
tools=[{
"type": "function",
"function": {
@ -694,7 +699,7 @@ async def test_named_tool_use(client: openai.AsyncOpenAI,
stream = await client.chat.completions.create(
model=MODEL_NAME,
messages=messages,
max_tokens=1000,
max_completion_tokens=1000,
tools=[{
"type": "function",
"function": {
@ -750,7 +755,7 @@ async def test_required_tool_use_not_yet_supported(
await client.chat.completions.create(
model=MODEL_NAME,
messages=messages,
max_tokens=1000,
max_completion_tokens=1000,
tools=[{
"type": "function",
"function": {
@ -765,7 +770,7 @@ async def test_required_tool_use_not_yet_supported(
await client.chat.completions.create(
model=MODEL_NAME,
messages=messages,
max_tokens=1000,
max_completion_tokens=1000,
tools=[{
"type": "function",
"function": {
@ -796,7 +801,7 @@ async def test_inconsistent_tool_choice_and_tools(client: openai.AsyncOpenAI,
with pytest.raises(openai.BadRequestError):
await client.chat.completions.create(model=MODEL_NAME,
messages=messages,
max_tokens=1000,
max_completion_tokens=1000,
tool_choice={
"type": "function",
"function": {
@ -809,7 +814,7 @@ async def test_inconsistent_tool_choice_and_tools(client: openai.AsyncOpenAI,
await client.chat.completions.create(
model=MODEL_NAME,
messages=messages,
max_tokens=1000,
max_completion_tokens=1000,
tools=[{
"type": "function",
"function": {

38
tests/entrypoints/openai/test_vision.py
View File

@ -78,11 +78,12 @@ async def test_single_chat_session_image(client: openai.AsyncOpenAI,
}]
# test single completion
chat_completion = await client.chat.completions.create(model=model_name,
messages=messages,
max_tokens=10,
logprobs=True,
top_logprobs=5)
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_completion_tokens=10,
logprobs=True,
top_logprobs=5)
assert len(chat_completion.choices) == 1
choice = chat_completion.choices[0]
@ -101,7 +102,7 @@ async def test_single_chat_session_image(client: openai.AsyncOpenAI,
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
)
message = chat_completion.choices[0].message
assert message.content is not None and len(message.content) >= 0
@ -134,7 +135,7 @@ async def test_single_chat_session_image_beamsearch(client: openai.AsyncOpenAI,
model=model_name,
messages=messages,
n=2,
max_tokens=10,
max_completion_tokens=10,
logprobs=True,
top_logprobs=5,
extra_body=dict(use_beam_search=True))
@ -169,11 +170,12 @@ async def test_single_chat_session_image_base64encoded(
}]
# test single completion
chat_completion = await client.chat.completions.create(model=model_name,
messages=messages,
max_tokens=10,
logprobs=True,
top_logprobs=5)
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_completion_tokens=10,
logprobs=True,
top_logprobs=5)
assert len(chat_completion.choices) == 1
choice = chat_completion.choices[0]
@ -192,7 +194,7 @@ async def test_single_chat_session_image_base64encoded(
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
)
message = chat_completion.choices[0].message
assert message.content is not None and len(message.content) >= 0
@ -226,7 +228,7 @@ async def test_single_chat_session_image_base64encoded_beamsearch(
model=model_name,
messages=messages,
n=2,
max_tokens=10,
max_completion_tokens=10,
extra_body=dict(use_beam_search=True))
assert len(chat_completion.choices) == 2
assert chat_completion.choices[
@ -259,7 +261,7 @@ async def test_chat_streaming_image(client: openai.AsyncOpenAI,
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
temperature=0.0,
)
output = chat_completion.choices[0].message.content
@ -269,7 +271,7 @@ async def test_chat_streaming_image(client: openai.AsyncOpenAI,
stream = await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
temperature=0.0,
stream=True,
)
@ -320,7 +322,7 @@ async def test_multi_image_input(client: openai.AsyncOpenAI, model_name: str,
await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
temperature=0.0,
)
@ -337,7 +339,7 @@ async def test_multi_image_input(client: openai.AsyncOpenAI, model_name: str,
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
max_completion_tokens=10,
temperature=0.0,
)
message = chat_completion.choices[0].message