Skip to content

GitLab

Commit 9b78a6ee authored by OlivierDehaene's avatar OlivierDehaene
Browse files

fix: only keep stop sequence buffer if we have some

parent 80a69204
Show whitespace changes
Inline Side-by-side
Showing with 10 additions and 9 deletions
server/text_generation_server/utils/tokens.py
View file @ 9b78a6ee
...@@ -112,7 +112,7 @@ class StoppingCriteria: ...@@ -112,7 +112,7 @@ class StoppingCriteria:
self.stop_sequence_criterias = stop_sequence_criterias self.stop_sequence_criterias = stop_sequence_criterias
self.max_new_tokens = max_new_tokens self.max_new_tokens = max_new_tokens
self.current_tokens = 0 self.current_tokens = 0
self.current_output = "test" self.current_output = ""
self.ignore_eos_token = ignore_eos_token self.ignore_eos_token = ignore_eos_token
def __call__(self, last_token: int, last_output: str) -> Tuple[bool, Optional[str]]: def __call__(self, last_token: int, last_output: str) -> Tuple[bool, Optional[str]]:
...@@ -123,6 +123,7 @@ class StoppingCriteria: ...@@ -123,6 +123,7 @@ class StoppingCriteria:
if not self.ignore_eos_token and last_token == self.eos_token_id: if not self.ignore_eos_token and last_token == self.eos_token_id:
return True, FinishReason.FINISH_REASON_EOS_TOKEN return True, FinishReason.FINISH_REASON_EOS_TOKEN
if self.stop_sequence_criterias:
self.current_output += last_output self.current_output += last_output
# There is no need to keep an output that is too long # There is no need to keep an output that is too long
if len(self.current_output) > 300: if len(self.current_output) > 300:
......
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