uclaml/SPPO

Issues

• Added a new PR to allow generation on fewer than 8 GPUs

  #25 opened 13 days ago by aman2304
  0

• Typo here

  #23 opened 21 days ago by xukp20
  1

• Scores and probability calcuations

  #15 opened 2 months ago by namdw
  3

• Dataset used and results in Gemma-2-9B results

  #12 opened 2 months ago by hodachi-axcxept
  13

• DPO baseline implementation

  #22 opened 2 months ago by yesiam-png
  0

• Is it normal the pipeline start with a huge loss ?

  #8 opened 2 months ago by qy1026
  3

• SPPO Implementation on Axolotl!

  #21 opened 2 months ago by kaykyr
  0

• Adaptation for 4-bit Quantization Training/Responses Generation (with 2 Home GPUs)

  #16 opened 2 months ago by kaykyr
  1

• What's the package configuration for reproduce SPPO-Gemma-2?

  #14 opened 2 months ago by Jackory
  1

• Any chance it work on my homelab?

  #13 opened 2 months ago by kaykyr
  3

• Suggestion: Gemma 2 9B and 27B.

  #3 opened 2 months ago by kaykyr
  2

• Which version of vllm should be installed

  #5 opened 2 months ago by xinghuang2050
  4

• Ranking speed & training hyperparameters

  #10 opened 2 months ago by skramer-dev
  0

• Some packages' version are too old

  #7 opened 2 months ago by qy1026
  0

• Questions about the training code

  #6 opened 2 months ago by blackblue9
  1

• ShareGPT appending

  #4 opened 2 months ago by Kquant03
  0

• Is it possible to run llama 3-70B and/or mixtral 8x22b through this process?

  #1 opened 3 months ago by RandomInternetPreson
  1

• ConnectionError: Couldn't reach 'synthetic_data_llama-3-8b-instruct-sppo-iter3_score' on the Hub (ConnectionError)

  #2 opened 3 months ago by xinghuang2050
  2