voidism/DoLa

Issues

• two questions about get_relative_top_filter?

  #20 opened a month ago by haiduo
  0

• Do you have the code for the visualization?

  #19 opened 3 months ago by JiaxinQin0814
  0

• concerns about your experiments of TruthfulQA-MC

  #7 opened a year ago by 0-KaiKai-0
  2

• The way to calculate the log_probs

  #18 opened 7 months ago by laurenlong
  0

• data for gpt-judge training

  #17 opened 7 months ago by itsmemala
  1

• Support for LLaMA-2

  #4 opened a year ago by ayyyq
  15

• Typo? mode = "early_exit_contrastive" -> mode = "dola-static"

  #16 opened 8 months ago by young-j-park
  1

• Some questions about the idea of the paper

  #6 opened 8 months ago by Rh-Dang
  1

• How to install the environment if I don't have root permission?

  #15 opened 8 months ago by hummingbird2030
  0

• Query regarding JS Divergence mean over batches.

  #13 opened 9 months ago by talha1503
  2

• What tool do you use to get the token prediction of each layer of large language models for Figure 2?

  #12 opened 9 months ago by frankdarkluo
  3

• dola可以适配xverse模型吗？

  #2 opened 10 months ago by xxm1668
  1

• Isn't the range of the JS divergence [0,1]? Why are there values greater than 1 in Figure 2?

  #10 opened a year ago by Jeryi-Sun
  1

• questions about code

  #3 opened a year ago by Cheung-Z
  1

• Should apply model.norm layer to hidden_states[early_exit_layer] ?

  #9 opened a year ago by githubhyz
  1

• Figure 2 subtoken "_"

  #5 opened a year ago by githubhyz
  0