Skip to content

[Frontend] Add Usage data in each chunk for chat_serving. #6540 #6652

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 8 commits into from
Jul 23, 2024
Merged

[Frontend] Add Usage data in each chunk for chat_serving. #6540 #6652

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
f78a2e3
change serving_chat to return usage data on each chunk of stream
Jul 22, 2024
49294b8
run formatter
Jul 22, 2024
f99de0e
add test for new continious_usage_stats option
Jul 23, 2024
57842b8
fix ruff
Jul 23, 2024
bda75b1
yapf fixer
Jul 23, 2024
8ec181a
should add continuous_stats to test for only include_usage
Jul 23, 2024
6b0d43d
simplify code
Jul 23, 2024
d87fe88
simplify code and run yapf
Jul 23, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
40 changes: 32 additions & 8 deletions tests/entrypoints/openai/test_chat.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -295,14 +295,19 @@ async def test_chat_completion_stream_options(client: openai.AsyncOpenAI,
async for chunk in stream:
assert chunk.usage is None

# Test stream=True, stream_options={"include_usage": True}
stream = await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
temperature=0.0,
stream=True,
stream_options={"include_usage": True})
# Test stream=True, stream_options={"include_usage": True,
# "continuous_usage_stats": False}}
stream = await client.chat.completions.create(model=model_name,
messages=messages,
max_tokens=10,
temperature=0.0,
stream=True,
stream_options={
"include_usage":
True,
"continuous_usage_stats":
False
})

async for chunk in stream:
if chunk.choices[0].finish_reason is None:
Expand Down Expand Up @@ -338,6 +343,25 @@ async def test_chat_completion_stream_options(client: openai.AsyncOpenAI,
stream=False,
stream_options={"include_usage": True})

# Test stream=True, stream_options={"include_usage": True,
# "continuous_usage_stats": True}
stream = await client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=10,
temperature=0.0,
stream=True,
stream_options={
"include_usage": True,
"continuous_usage_stats": True
},
)
async for chunk in stream:
assert chunk.usage.prompt_tokens >= 0
assert chunk.usage.completion_tokens >= 0
assert chunk.usage.total_tokens == (chunk.usage.prompt_tokens +
chunk.usage.completion_tokens)


# NOTE: Not sure why, but when I place this after `test_guided_regex_chat`
# (i.e. using the same ordering as in the Completions API tests), the test
Expand Down
50 changes: 46 additions & 4 deletions vllm/entrypoints/openai/serving_chat.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -222,7 +222,15 @@ async def chat_completion_stream_generator(
model=model_name)
if (request.stream_options
and request.stream_options.include_usage):
chunk.usage = None
if (request.stream_options.continuous_usage_stats):
prompt_tokens = len(res.prompt_token_ids)
usage = UsageInfo(prompt_tokens=prompt_tokens,
completion_tokens=0,
total_tokens=prompt_tokens)
chunk.usage = usage
else:
chunk.usage = None

data = chunk.model_dump_json(exclude_unset=True)
yield f"data: {data}\n\n"

Expand Down Expand Up @@ -252,7 +260,18 @@ async def chat_completion_stream_generator(
model=model_name)
if (request.stream_options and
request.stream_options.include_usage):
chunk.usage = None
if (request.stream_options.
continuous_usage_stats):
prompt_tokens = len(
res.prompt_token_ids)
usage = UsageInfo(
prompt_tokens=prompt_tokens,
completion_tokens=0,
total_tokens=prompt_tokens)
chunk.usage = usage
else:
chunk.usage = None

data = chunk.model_dump_json(
exclude_unset=True)
yield f"data: {data}\n\n"
Expand Down Expand Up @@ -311,7 +330,19 @@ async def chat_completion_stream_generator(
model=model_name)
if (request.stream_options
and request.stream_options.include_usage):
chunk.usage = None
if (request.stream_options.continuous_usage_stats):
prompt_tokens = len(res.prompt_token_ids)
completion_tokens = len(output.token_ids)
usage = UsageInfo(
prompt_tokens=prompt_tokens,
completion_tokens=completion_tokens,
total_tokens=prompt_tokens +
completion_tokens,
)
chunk.usage = usage
else:
chunk.usage = None

data = chunk.model_dump_json(exclude_unset=True)
yield f"data: {data}\n\n"
else:
Expand All @@ -331,7 +362,18 @@ async def chat_completion_stream_generator(
model=model_name)
if (request.stream_options
and request.stream_options.include_usage):
chunk.usage = None
if (request.stream_options.continuous_usage_stats):
prompt_tokens = len(res.prompt_token_ids)
completion_tokens = len(output.token_ids)
usage = UsageInfo(
prompt_tokens=prompt_tokens,
completion_tokens=completion_tokens,
total_tokens=prompt_tokens +
completion_tokens,
)
chunk.usage = usage
else:
chunk.usage = None
data = chunk.model_dump_json(exclude_unset=True)
yield f"data: {data}\n\n"
finish_reason_sent[i] = True
Expand Down