Xie Yuhang

代码疑问：矩阵计算batch内的cos loss

1

为什么“取出负例-正例的差值”这里用

模型微调问题

3

您好！想用自己的jsonl文件，微调text2vec-base-multilingual模型，用python training_sup_text_matching_model_jsonl_data.py --model_arch cosent --do_train --do_predict --num_epochs 10 指令。会出现如下报错，OSError: ./outputs/jsonl-model does not appear to have a file named config.json. Checkout 'https://huggingface.co/./outputs/jsonl-model/None' for available files.。请问应该怎么修改指令或代码。 

Could you please provide the formulas for Rouge_1_r, Rouge_1_p, Rouge_1_f, Rouge_2_r, Rouge_2_p, Rouge_2_f, Rouge_l_r, Rouge_l_p, and Rouge_l_f separately

NotImplementedError: Loading a dataset cached in a LocalFileSystem is not supported.解决后，又遇到datasets会和rl4lms冲突

9

遇到了一个报错NotImplementedError: Loading a dataset cached in a LocalFileSystem is not supported.然后我看了一下github上的issue，好像都是这几天才发现的问题。按照github上的解决方法，是更新datasets包。但是更新datasets会和rl4lms冲突，rl4lms会退版本，然后报几个变量定义为空的错；更新rl4lms会把datasets退版本，然后报错: LocalFileSystem is not supported.还有一种方法是升级这个包，pip install fsspec==2023.9.2，但是也是会报错变量未定义。未定义的变量是selected_data和test_data等。  

An error was encountered “NotImplementedError: Loading a dataset cached in a LocalFileSystem is not supported”. Then I looked at the issues on github, and it seemed like they had only...