12B模型ValueError: Tokenizer class TelechatTokenizer does not exist or is not currently imported.

[wuxiulike]

运行python telechat_service.py 报错 Traceback (most recent call last): File "/root/Telechat/service/telechat_service.py", line 16, in tokenizer = AutoTokenizer.from_pretrained(PATH) File "/root/anaconda3/envs/telechat/lib/python3.10/site-packages/transformers/models/auto/tokenization_auto.py", line 688, in from_pretrained raise ValueError( ValueError: Tokenizer class TelechatTokenizer does not exist or is not currently imported. (telechat) root@estar-ESC8000-G4:~/Telechat/service# python telechat_service.py Traceback (most recent call last): File "/root/Telechat/service/telechat_service.py", line 16, in tokenizer = AutoTokenizer.from_pretrained(PATH) File "/root/anaconda3/envs/telechat/lib/python3.10/site-packages/transformers/models/auto/tokenization_auto.py", line 688, in from_pretrained raise ValueError( ValueError: Tokenizer class TelechatTokenizer does not exist or is not currently imported.

模型为7B时可以正常运行，修改为12B路径时，报上面的错

###############telechat_service.py#################### import uvicorn import os

from fastapi.middleware.cors import CORSMiddleware from contextlib import asynccontextmanager from fastapi.encoders import jsonable_encoder from fastapi import FastAPI from fastapi.responses import StreamingResponse

import torch from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig

os.environ["CUDA_VISIBLE_DEVICES"] = "0" PATH = '/opt/models/TeleChat-12B' tokenizer = AutoTokenizer.from_pretrained(PATH) model = AutoModelForCausalLM.from_pretrained(PATH, trust_remote_code=True, device_map="auto", torch_dtype=torch.float16) generate_config = GenerationConfig.from_pretrained(PATH) model.eval()

[wuxiulike]

找到问题了，把telechat_service.py第16行 tokenizer = AutoTokenizer.from_pretrained(PATH) 修改为： tokenizer = AutoTokenizer.from_pretrained(PATH,trust_remote_code=True) 问题解决。

[ckcoding]

感谢，我也遇到了你的这个问题，按你的已解决