Equationliu/Kangaroo

Issues

• Kangaroo when bsz is greater than 1.

  #6 opened 2 months ago by cool-xiang
  2

• Encountering NaN output at a specific batch ID every run, and no change observed upon adjusting the learning rate

  #5 opened 3 months ago by Zerohclmax
  0

• a question

  #4 opened 4 months ago by cool-xiang
  2

• In line 263 of train.py， predict = model(inputs_embeds=data["hidden_states_early"]

  #3 opened 4 months ago by cool-xiang
  1

• Training procedure of Kangaroo.

  #2 opened 4 months ago by tim-pan
  5

• why warmup when evaluating

  #1 opened 6 months ago by EganGu
  2