Well, it seems that hugging-face has rebuilt their repository and in your ReadMe.md, you just provide the link to the original position of run_glue.py to tune the teacher. Is there any possibility that you might provide the target version of run_glue.py?

Hi thanks for your work.
I am studying LoRA and I noticed that the your work Dylora dose not contain the implementation of NLG part.
I am simply writing to ask in order to reproduce the results of generative tasks, should I change the hyper-parameter lora-dim in as you did with lora_r in the NLU task?

hi, the mate-kd is an excellent work on NLP KD. Here I have a question about the codes of this paper.

In the section 4.1 of the paper, the authors said that two different teacher models (Roberta large and BERT base) were used in the two steps, but the codes showed that only one teacher model is used. Is it right?

on the other hand, the two steps should be trained separately? But the codes showed that in the training procedure, 10 steps for updating the params of generator, then 100 steps for updating the student model. That makes me feel wired.

Keep getting this file not found error for DyLora base model. Not sure where to find it.

When I run DyLoRA‘s cola task, I get this error.

Hi, thanks for sharing your code of this amazing paper!

I have a probably naive question regarding the LoRA baseline implementation. How did you find the low-rank matrix in table to accommodate small-device deployment? Was it done by selecting the first r vectors in LoRA?

Thanks for your response in advance.