Hao

icon company Kriston AI icon location RA at CUHK, SZ

Results 2 issues of Hao

questions about the padding value

1 comment

https://github.com/IDEA-CCNL/Fengshenbang-LM/blob/c8fb7b8437843ea13fa9d147ce86c4592fa21237/fengshen/examples/qa_t5/qa_dataset.py#L97-L98 https://github.com/IDEA-CCNL/Fengshenbang-LM/blob/c8fb7b8437843ea13fa9d147ce86c4592fa21237/fengshen/examples/qa_t5/qa_dataset.py#L115-L121 First of all, thank you for your code. Your code helps me a lot. I have a small question on how you pad the input sequences. In Lines...

Is this line wrong?

https://github.com/liucongg/ChatGLM-Finetuning/blob/d52202a2facfa1dff45b0daec7e56aa54c126616/utils.py#L95-L96 I am not very sure. But this seems the right code: ` input_ids = tokenizer.convert_tokens_to_ids(tokens) + [tokenizer.get_command("[gMASK]"), tokenizer.get_command("sop")] `