Open hzhaoy opened 3 months ago
操作系统:Ubuntu 22.04.3 LTS GPU:NVIDIA A100-SXM4-80GB CUDA:Driver Version: 535.154.05 CUDA Version: 12.2
代码复现
import os import torch from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig os.environ["CUDA_VISIBLE_DEVICES"] = '0' PATH = '/models/Tele-AI/telechat-7B/' 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() inputs = "hello" print("输入:", inputs) output = model.generate(**tokenizer(inputs, return_tensors="pt").to(model.device), generation_config=generate_config) output = tokenizer.decode(output[0]) print("续写结果:", output) inputs = "你是" output = model.generate(**tokenizer(inputs, return_tensors="pt").to(model.device), generation_config=generate_config) output = tokenizer.decode(output[0]) print("续写结果:", output)
输出结果:
您确认一下使用的模型是base模型么?目前暂时没开放base模型,所以这个api一般情况下不会被调用才对。
不是base模型,是7B-Chat模型 执行inference_telechat/telechat_infer_demo.py中的这段代码
操作系统:Ubuntu 22.04.3 LTS GPU:NVIDIA A100-SXM4-80GB CUDA:Driver Version: 535.154.05 CUDA Version: 12.2
代码复现
输出结果:
![13372](https://github.com/Tele-AI/Telechat/assets/61868295/75ca565f-917b-460d-bc4f-809b45fb4a0a)