from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_name = "BlankZ/ragen-checkpoint-step-600-bf16"

model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

# 生成文本
# 注意：请根据您的训练任务调整prompt格式
messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "你好"}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

outputs = model.generate(**model_inputs, max_length=100)
print(tokenizer.decode(outputs[0]))

注意事项

这是RLHF训练后的Actor模型，用于文本生成。
使用BF16精度以节省显存，建议使用支持BF16的GPU (如A100, H100等)。