[BUG/Help] <title>本地调用模型失败，报错'ChatGLMTokenizer' object has no attribute 'tokenizer'

arx-night commented 1 year ago

Is there an existing issue for this?

[X] I have searched the existing issues

Current Behavior

就是采用最简单的chatGlm演示demo，然后发现报错 'ChatGLMTokenizer' object has no attribute 'tokenizer'

Expected Behavior

No response

Steps To Reproduce

就是采用最简单的demo，然后chatGLM模型是我这边从hugging face官网上手动下载的结果却报了这个错，是我这边下载的方式不对吗？还是模型有问题？

Environment

- OS:Ubuntu   20.04.5
- Python: 3.8.10
- Transformers: 4.26.1
- PyTorch: 1.12
- CUDA Support (`python -c "import torch; print(torch.cuda.is_available())"`) :True

Anything else?

No response

xieyongshuai commented 1 year ago

有解决吗？同问

shenlijungg commented 1 year ago

刚下载也是一样的问题

xieyongshuai commented 1 year ago

已解决，去huggface下最新的模型

shenlijungg commented 1 year ago

@xieyongshuai 我是清华的地址下载的，不能用吗

edisonwd commented 1 year ago

有解决吗？同问

如果 transformers==4.34.0 会报 'ChatGLMTokenizer' object has no attribute 'tokenizer' 解决方法是，降低 transformers 版本，安装下面的版本可以正常运行：

pip install transformers==4.33.2

THUDM / ChatGLM2-6B