l294265421/alpaca-rlhf

Issues

• deepspeed.initialize的一些疑惑

  #8 opened a year ago by iamsile
  8

• 增大max_prompt_len和max_ans_len训练会出现非法的内存访问问题

  #16 opened a year ago by Luoxiaohei41
  0

• 训练问题

  #15 opened a year ago by wanghao-007
  0

• Step 3: Actor model和Reward model使用不同的tokenizer

  #14 opened a year ago by Kevin-myxu
  0

• step2和step3中padding side似乎不一样？

  #13 opened a year ago by qiancheng99
  1

• A question about setting tokens

  #12 opened a year ago by hepj987
  1

• element 0 of tensors does not require grad and does not have a grad_fn

  #11 opened a year ago by Bill-Orz
  5

• v100 step3 oom

  #6 opened a year ago by iamsile
  12

• 关于Step3中是否需要把生成的answer中eos后面token mask掉

  #9 opened a year ago by Ablustrund
  1

• Fix pad_token_id bug

  #10 opened a year ago by Ablustrund
  2

• how to run it, need more details

  #7 opened a year ago by SeekPoint
  2

• stop at step2 evaluation_reward

  #5 opened 2 years ago by murphypei
  4

• 训练效果怎么样

  #1 opened 2 years ago by Curious-chen
  3

• reward model在v100上训练时会卡住不动

  #4 opened 2 years ago by iamsile
  2

• Steps

  #2 opened 2 years ago by syngokhan
  1

• v100训练时显存oom

  #3 opened 2 years ago by iamsile
  2