Skip to content

GitLab

Unverified Commit cb8ac1b0 authored by lvhan028's avatar lvhan028 Committed by GitHub
Browse files

fix crash when conversation history out of limit (#28)

parent c16b857b
Hide whitespace changes
Inline Side-by-side
Showing with 5 additions and 0 deletions
llmdeploy/serve/fastertransformer/chatbot.py
View file @ cb8ac1b0
...@@ -168,6 +168,8 @@ class Chatbot: ...@@ -168,6 +168,8 @@ class Chatbot:
sequence_start, sequence_start,
sequence_end): sequence_end):
yield status, res, tokens yield status, res, tokens
if status.value < 0:
return
self._session.histories = \ self._session.histories = \
self._session.histories + self._session.prompt + \ self._session.histories + self._session.prompt + \
self._session.response self._session.response
...@@ -332,7 +334,10 @@ class Chatbot: ...@@ -332,7 +334,10 @@ class Chatbot:
f'history tokens {session.sequence_length}, ' \ f'history tokens {session.sequence_length}, ' \
f'request length {request_output_len}' f'request length {request_output_len}'
yield StatusCode.TRITON_SESSION_OUT_OF_LIMIT, errmsg, 0 yield StatusCode.TRITON_SESSION_OUT_OF_LIMIT, errmsg, 0
return
logger.info(f'session {session.session_id}, ' logger.info(f'session {session.session_id}, '
f'max length: {self.cfg.session_len}, '
f'input tokens: {input_tokens}, ' f'input tokens: {input_tokens}, '
f'request tokens: {request_output_len}, ' f'request tokens: {request_output_len}, '
f'history tokens: {session.sequence_length}') f'history tokens: {session.sequence_length}')
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment