github
/
InternLM
mirror of https://github.com/InternLM/InternLM
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

Update modeling_internlm.py 

fixed the issue that the HF model spontaneously conducted multiple rounds of Q&A and stream_chat method generates garbled characters
pull/560/head
djsaber 2023-12-26 16:52:06 +08:00 committed by GitHub
parent 97e7d03d09
commit 508711cc97
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
tools/transformers/modeling_internlm.py
View File

@ -861,7 +861,7 @@ class InternLMForCausalLM(InternLMPreTrainedModel):
self.chche.extend(value.tolist())
token = self.tokenizer.decode(self.chche, skip_special_tokens=True)
if " " in token and len(token) <= 5:
if "<EFBFBD>" in token and len(token) <= 5:
return
if token.strip() != "<eoa>":