环境:使用 AutoTokenizer 加载量化后的百川模型
原始加载分词代码

from transformers import AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained(
    model_id,
    torch_dtype=torch.float32,
    use_fast=False
)

报错:ValueError: Tokenizer class BaichuanTokenizer does not exist or is not currently imported

解决代码:

from transformers import AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained(
    model_id,
    torch_dtype=torch.float32,
    use_fast=False, 
    trust_remote_code=True
)

添加 trust_remote_code=True

可以在量化时对 Tokenizer 进行保存

tokenizer.save_pretrained('YOUR_PATH')
Logo

一站式 AI 云服务平台

更多推荐