DeepKE icon indicating copy to clipboard operation
DeepKE copied to clipboard

baichuan2-13b-iepile-lora模型预测报错zjunlp/baichuan2-13b-iepile-lora does not appear to have a file named config.json.

Open Funnyxiaxia opened this issue 1 year ago • 3 comments

在案例:Lora微调Baichuan完成CCKS2023指令驱动的知识图谱构建 中,我尝试使用训练好的baichuan2-13b-iepile-lora模型进行预测,这是我的bash文件:CUDA_VISIBLE_DEVICES=0 python src/inference.py \ --stage sft \ --model_name_or_path 'zjunlp/baichuan2-13b-iepile-lora' \ --checkpoint_dir 'lora/baichuan2-13b-iepile-lora' \ --model_name 'baichuan' \ --template 'baichuan2' \ --do_predict \ --input_file 'data/input.json' \ --output_file 'results/baichuan2-13b-iepile-lora_output.json' \ --finetuning_type lora \ --output_dir 'lora/test' \ --predict_with_generate \ --cutoff_len 512 \ --bf16 \ --max_new_tokens 300 \ --bits 4,执行后报错OSError: zjunlp/baichuan2-13b-iepile-lora does not appear to have a file named config.json. Checkout 'https://huggingface.co/zjunlp/baichuan2-13b-iepile-lora/tree/main' for available files.

Funnyxiaxia avatar May 19 '24 18:05 Funnyxiaxia

你好,model_name_or_path 是底座模型即https://huggingface.co/baichuan-inc/Baichuan2-13B-Chat。 另外,我们新发布的信息抽取大模型OneKE在信息抽取方面效果更佳 https://github.com/zjunlp/DeepKE/blob/main/example/llm/OneKE.md

guihonghao avatar May 20 '24 01:05 guihonghao

我将--model_name_or_path 'baichuan-inc/Baichuan2-13B-Chat' \更换为'baichuan-inc/Baichuan2-13B-Chat'后运行,报错ValueError: You can't pass load_in_4bitor load_in_8bit as a kwarg when passing quantization_config argument at the same time.

Funnyxiaxia avatar May 20 '24 02:05 Funnyxiaxia

你好,我们采用的环境是

accelerate==0.21.0
transformers==4.33.0
bitsandbytes==0.39.1

guihonghao avatar May 20 '24 07:05 guihonghao

您好,请问您的问题是否已解决?

zxlzr avatar May 24 '24 03:05 zxlzr

您好,使用了OneKE。

Funnyxiaxia avatar May 24 '24 04:05 Funnyxiaxia

能否详述一下example/llm/InstructKGC的评估脚本,在使用OneKE继续训练后如何使用测试集,又如何使用评估脚本。

Funnyxiaxia avatar May 24 '24 17:05 Funnyxiaxia

1、参考 data 目录下各个任务中sample.json文件格式组织测试文件,schema.json组织schema信息。 2、参考 https://github.com/zjunlp/DeepKE/blob/main/example/llm/InstructKGC/README_CN.md#23%E6%B5%8B%E8%AF%95%E6%95%B0%E6%8D%AE%E8%BD%AC%E6%8D%A2 步骤,按照下面的代码转换sample.json文件为模型可输入的文件test.json

python ie2instruction/convert_func.py \
    --src_path data/NER/sample.json \
    --tgt_path data/NER/test.json \
    --schema_path data/NER/schema.json \
    --language zh \
    --task NER \
    --split_num 6 \
    --split test

转换后的每条测试数据将含有id, instruction, label字段。 3、输入模型,模型预测输出output字段。 4、通过下面的代码计算F1分数。

python ie2instruction/eval_func.py \
  --path1 '模型输出文件路径' \
  --task NER 

guihonghao avatar May 25 '24 03:05 guihonghao

在NER任务中,假正例、假负例、真负例如何解释

Funnyxiaxia avatar May 25 '24 03:05 Funnyxiaxia

参考代码:https://github.com/zjunlp/DeepKE/blob/main/example/llm/InstructKGC/ie2instruction/eval/metric/ner_metric.py

guihonghao avatar May 25 '24 03:05 guihonghao

pred_num和gold_num是什么意思

Funnyxiaxia avatar May 25 '24 04:05 Funnyxiaxia

参考资料:https://blog.csdn.net/weixin_45498383/article/details/135740711

guihonghao avatar May 25 '24 04:05 guihonghao

十分感谢

Funnyxiaxia avatar May 25 '24 04:05 Funnyxiaxia

请问您还有其他问题吗?

zxlzr avatar May 25 '24 06:05 zxlzr