Skip to content
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

Jump to bottom

fix swift deploy (lmdeploy stream) #2397

Merged
merged 2 commits into from
Nov 6, 2024
Merged

fix swift deploy (lmdeploy stream) #2397

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
1ce44c5
fix swift deploy (lmdeploy stream)
Jintao-Huang Nov 6, 2024
8d3d4b0
update
Jintao-Huang Nov 6, 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
20 changes: 6 additions & 14 deletions swift/llm/deploy.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -516,9 +516,7 @@ async def _generate_full():
prompt_tokens=num_prompt_tokens,
completion_tokens=num_generated_tokens,
total_tokens=num_prompt_tokens + num_generated_tokens)
finish_reason = None
if output.status.name == 'FINISH':
finish_reason = 'stop'
finish_reason = 'stop' if output.status.name == 'FINISH' else None

if isinstance(request, ChatCompletionRequest):
action, action_input = split_action_action_input(response)
Expand Down Expand Up @@ -554,31 +552,25 @@ async def _generate_stream():
async with llm_engine.safe_run(session_id):
async_iter = generator.async_stream_infer(
session_id=session_id, **inputs, stream_output=True, gen_config=generation_config).__aiter__()
is_finished = False
response = None
while not is_finished:
try:
output = await async_iter.__anext__()
except StopAsyncIteration:
is_finished = True
async for output in async_iter:
num_generated_tokens = len(output.token_ids)
usage_info = UsageInfo(
prompt_tokens=num_prompt_tokens,
completion_tokens=num_generated_tokens,
total_tokens=num_prompt_tokens + num_generated_tokens,
)
is_finished = output.status.name == 'FINISH'
delta_text = template.generate_ids_to_response(
output.token_ids, is_finished, return_delta=True, print_idx=print_idx)

finish_reason = None
if output.status.name == 'FINISH':
finish_reason = 'stop'
if not delta_text and finish_reason != 'stop':
if not delta_text and not is_finished:
continue
finish_reason = 'stop' if is_finished else None
total_response += delta_text
if isinstance(request, ChatCompletionRequest):
toolcall = None
if finish_reason == 'stop':
if is_finished:
action, action_input = split_action_action_input(total_response)
if action is not None:
toolcall = [
Expand Down
Loading