LaVieEnRose365 / ReLLa

Code of Paper "ReLLa: Retrieval-enhanced Large Language Models for Mitigating Long Context Problems in Recommendation".
34 stars 4 forks source link

get_semantic_embed 出错 #12

Open lightningsoon opened 1 week ago

lightningsoon commented 1 week ago

python get_semantic_embed.py --model_path ./Llama-2-7b-hf --dataset BookCrossing --pooling average --gpu_id 1 miniconda3/envs/rella/lib/python3.10/site-packages/transformers/configuration_utils.py:902 in dict_torch_dtype_to_str │ │ │ │ 899 │ │ string, which can then be stored in the json format. │ │ 900 │ │ """ │ │ 901 │ │ if d.get("torch_dtype", None) is not None and not isinstance(d["torch_dtype"], s │ │ ❱ 902 │ │ │ d["torch_dtype"] = str(d["torch_dtype"]).split(".")[1] │ │ 903 │ │ for value in d.values(): │ │ 904 │ │ │ if isinstance(value, dict): │ │ 905 │ │ │ │ self.dict_torch_dtype_to_str(value)

有两个问题 1、用哪个embedding model 2、模型载入就出错了,看起来像是版本问题

LaVieEnRose365 commented 1 week ago

我们论文中使用的是Vicuna-13b-v1.3,对于第二个问题,这个报错我们之前没有遇到过,能对应到代码哪一行报错的吗