NVlabs/DoRA

Issues

• Save model checkpoint error when multi-gpu training

  #18 opened a month ago by azuryl
  1

• Unable to find commonsense_170k.json'

  #17 opened a month ago by azuryl
  0

• Cannot reproduce the results of Llama7B dora_r32.

  #14 opened 3 months ago by xiaoshingshing2
  7

• Hi, thanks for you code，but I got very low accuray when used the lora as fllow.

  #16 opened 2 months ago by LiZhangMing
  0

• Can you provide the code to reproduce Fig2？

  #15 opened 3 months ago by 2018211801
  0

• What is the lora_alpha used in commonsense_reasoning?

  #13 opened 3 months ago by xiaoshingshing2
  2

• About the dimension discrepancy of the magnitude vector

  #11 opened 4 months ago by Swiminn
  4

• Cannot reproduce the accuracy of llama2-7b on CS using the llama2_7B_DoRA.sh script.

  #9 opened 4 months ago by LuletterSoul
  18

• CANT import DoraConfig from peft ?

  #10 opened 4 months ago by Cheng-lg
  0

• When will Hugging Face support QDORA?

  #8 opened 4 months ago by 190679163
  1

• About the training seed

  #2 opened 5 months ago by lucasliunju
  4

• RuntimeError: Internal: src/sentencepiece_processor.cc(1101) [model_proto->ParseFromArray(serialized.data(), serialized.size())]

  #6 opened 4 months ago by 2tianyao1
  2

• Support for the `Embedding` module

  #1 opened 5 months ago by ShayekhBinIslam
  1

• Using LLaMA-3-8b in visual instruction tuning

  #3 opened 5 months ago by enkaranfiles
  1

• OOM when use torch.linalg.norm

  #4 opened 5 months ago by shuishen112
  1