whybeyoung
/
chatglm-6b-int4
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix logit processor 

Fix tokenizer config saving
This commit is contained in:
duzx16 2023-04-08 12:07:06 +08:00
parent 96de7a2fb3
commit a7272d4c93
2 changed files with 10 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
modeling_chatglm.py
View File

@ -56,7 +56,7 @@ class InvalidScoreLogitsProcessor(LogitsProcessor):
def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor) -> torch.FloatTensor: def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor) -> torch.FloatTensor:
if torch.isnan(scores).any() or torch.isinf(scores).any(): if torch.isnan(scores).any() or torch.isinf(scores).any():
scores.zero_() scores.zero_()
scores[..., 20005] = 5e4 scores[..., 5] = 5e4
return scores return scores

12
tokenization_chatglm.py
View File

@ -170,9 +170,9 @@ class ChatGLMTokenizer(PreTrainedTokenizer):
vocab_file, vocab_file,
do_lower_case=False, do_lower_case=False,
remove_space=False, remove_space=False,
bos_token='sop', bos_token='<sop>',
eos_token='eos', eos_token='</s>',
eop_token='eop', eop_token='<eop>',
mask_token='[MASK]', mask_token='[MASK]',
gmask_token='[gMASK]', gmask_token='[gMASK]',
padding_side="left", padding_side="left",
@ -183,6 +183,12 @@ class ChatGLMTokenizer(PreTrainedTokenizer):
do_lower_case=do_lower_case, do_lower_case=do_lower_case,
remove_space=remove_space, remove_space=remove_space,
padding_side=padding_side, padding_side=padding_side,
bos_token=bos_token,
eos_token=eos_token,
eop_token=eop_token,
mask_token=mask_token,
gmask_token=gmask_token,
num_image_tokens=num_image_tokens,
**kwargs **kwargs
) )