Open Coobiw opened 1 year ago
Hello, thanks to your great work!
In blip2_vicuna_instruct.py, the bos_token of LLM has been changed. Originally, it is '< s >' with idx:1. But after the following code:
blip2_vicuna_instruct.py
bos_token
self.llm_tokenizer.add_special_tokens({'pad_token': '[PAD]'}) self.llm_tokenizer.add_special_tokens({'bos_token': '</s>'}) self.llm_tokenizer.add_special_tokens({'eos_token': '</s>'}) self.llm_tokenizer.add_special_tokens({'unk_token': '</s>'})
After this ops, the bos_token is changed into '< /s >' with idx:2, the same as the eos_token.
eos_token
Is it a writting Error? Had you used this code for instruction tuning? If used, will it lead some bad influences?
Thanks for your reply!
hello, i find it too. Have you solved the problem?
It's been a long time. I remember that I've modify the token_id setting in generate function of InstructBLIP. Actually, this doesn't have much impact at all.
generate
Hello, thanks to your great work!
In
blip2_vicuna_instruct.py
, thebos_token
of LLM has been changed. Originally, it is '< s >' with idx:1. But after the following code:After this ops, the
bos_token
is changed into '< /s >' with idx:2, the same as theeos_token
.Is it a writting Error? Had you used this code for instruction tuning? If used, will it lead some bad influences?
Thanks for your reply!