Qwen
/

Text Generation
Transformers
Safetensors
Chinese
English
qwen
custom_code

Attempt to add in add_ Special_ Tokens, there is an error, how should I solve it

#19
by wangchengfei - opened

Here is my code section

tokenizer=AutoTokenizer.from_pretrained("Qwen/Qwen-7B", use_fast=False,trust_remote_code=True)
tokenizer.add_special_tokens({"pad_token":'[PAD]'})

But it made the following error:
File "Qwen-7B/tokenization_qwen.py", line 165, in _add_tokens raise ValueError("Adding unknown special tokens is not supported") ValueError: Adding unknown special tokens is not supported

Qwen org

We have a custom check in the tokenzier. Please try using <|extra_0|> and the like. Setting pad_token to <|endoftext|> should be generally okay for the pretrained models.

Sign up or log in to comment