[论文笔记] 检查LLM模型输出是否正常

model_test.py 

from transformers import AutoTokenizer, AutoModelForCausalLM
import warnings
import os
warnings.filterwarnings("ignore")

hf_path_list=["hf_qwen2_1b"]

for hf_path in hf_path_list:
    print("[M.M]", hf_path)
    # 加载分词器和模型
    tokenizer = AutoTokenizer.from_pretrained(hf_path)
    model = AutoModelForCausalLM.from_pretrained(hf_path)

    # 编码输入文本并生成注意力掩码
    # input_text = "Hey, are you conscious? Can you talk to me?"
    input_text = "There was a thick forest on the sides of a mountain. Many kinds of animals lived in the forest."
    input_ids = tokenizer(input_text, return_tensors="pt").input_ids
    attention_mask = (input_ids != tokenizer.pad_t