Merge pull request #508 from ctejada85/main

Now the last token sent when `stream=True`
2023-07-20 16:07:54 -04:00 · 2023-07-20 16:07:54 -04:00 · 5549a1cabd
commit 5549a1cabd
parent 36872620d0 0756a2d3fb
1 changed files with 1 additions and 1 deletions
--- a/llama_cpp/llama.py
+++ b/llama_cpp/llama.py
@ -953,7 +953,7 @@ class Llama:
                    token_end_position += len(self.detokenize([token]))
                    # Check if stop sequence is in the token
                    if token_end_position >= (
-                        remaining_length - first_stop_position - 1
+                        remaining_length - first_stop_position
                    ):
                        break
                    logprobs_or_none: Optional[CompletionLogprobs] = None