qywu / chinese-gpt Goto Github PK

I firstly think "model_state_epoch_62.th" came from tutorials/Training.ipynb, but I just found it do load_state_dict inside. Seems I need do "Model Weights Conversion(*Optional).ipynb" first and then do "Training.ipynb". but there isn't a script to generate "model_state_epoch_62.th" in every scripts in tutorials. Can you give some brief introduce on hwo to generate model_state_epoch_62.th?

Hey I'm trying to do some fine tuning on my own dataset using the Training.ipynb, but I don't know the structure of the training data. Can you show me how the training data should look like?

can't open links in "colab demo" and "Decoder Weights" in readme.md. Also confused about "summarization", "origin_title", "third_title" in tutorials/Data Processing Example.ipynb means for?
Does it get from origin dataset such as "baike2018qa" directly and can you describ how to get this? thanks

如题，pip install chinese-gpt 如何安装这个要怎么安装呢，网上查了很多资料都找不到原因。
希望可以得到回复

pre-training训练量太大了，如果可以分享下模型参数真是太cool了