neelsjain/NEFTune

Issues

• {RecursionError}maximum recursion depth exceeded while calling a Python object

  #5 opened a year ago by NormXU
  3

• Unable to evaluate the model

  #9 opened 10 months ago by sglucas
  9

• [Reimplementation] Unable to reproduce results -- Training loss curves are similar

  #1 opened 10 months ago by maximegmd
  11

• Output 0 of ViewBackward0 is a view

  #10 opened 10 months ago by clechristophe
  3

• how to add it to transformers‘s mdoel??

  #11 opened 10 months ago by Dhaizei
  2

• No eval_scoring.py file in the repo

  #12 opened 10 months ago by cuongtran-uva
  3

• quesiton about the noise injection location

  #15 opened 10 months ago by zhhao1
  1

• how to add multi-turns function?

  #14 opened 10 months ago by Dhaizei
  1

• can you suggest about chat tuning？

  #13 opened a year ago by Dhaizei
  0

• Question about output embedding from noised tokens

  #7 opened a year ago by isamu-isozaki
  1

• i think "return model" should be within the scope of the NEFTune function, not outside it

  #8 opened a year ago by Kayce001
  1

• RuntimeError: ``sharded_state_dict`` can only be used when parameters are flatten and sharded.

  #2 opened a year ago by Sniper970119
  9

• More benchmarks

  #6 opened a year ago by eyuansu62
  1

• QLoRA implementation

  #3 opened a year ago by marianbastiUNRN
  1