RLHF-V/RLAIF-V

Issues

• dependency conflicting when installing RLAIF_V

  #40 opened 18 days ago by zwang-datascience
  1

• Self feedback data generation pipeline & reference model

  #6 opened 7 months ago by charismaticchiu
  9

• question about the Labeler selection

  #39 opened a month ago by lyklly
  0

• question about the training loss?

  #38 opened a month ago by lufanma
  4

• batch size=1 , only a Nvidia a100 memory=80, why display cuda out of memory?

  #22 opened 2 months ago by AderonHuang
  9

• Error when initing

  #34 opened a month ago by pspdada
  3

• Support for Retraining MiniCPM-V?

  #35 opened a month ago by DarioPTWR
  0

• vision_tower seem not load

  #26 opened 4 months ago by zhangzef
  3

• where to find the lora fine-tuned checkpoints?

  #32 opened a month ago by lufanma
  2

• Training bugs - UnboundLocalError: local variable 'df' referenced before assignment

  #29 opened 2 months ago by youthHan
  2

• The LoRA training codes and scripts

  #11 opened 2 months ago by darkpromise98
  2

• Issues for Implementation

  #30 opened 2 months ago by injadlu
  3

• Inconsistent Training Parameters in 'trainer_state' and Reproduction Issues

  #27 opened 2 months ago by Molly-3000
  1

• about the RefoMB evaluation experiment

  #24 opened 2 months ago by Molly-3000
  1

• Follow the setting in the paper, but I can't reproduce the results.

  #28 opened 2 months ago by RobitsG
  4

• Questions Regarding the Training Data and Settings for LLaVA as Used in the Paper

  #15 opened 5 months ago by Timsty1
  4

• Would the data generation code be released?

  #21 opened 4 months ago by Gaffey
  1

• the actual number of samples of the huggingface RLAIF-V-Dataset is 83k, not 30k?

  #25 opened 4 months ago by Molly-3000
  2

• When I use the cal_logp of all dataset,I met the question.This

  #20 opened 4 months ago by XiaoLei2123
  4

• 我想问下 数据中logps怎么来的

  #16 opened 4 months ago by Spring24ch
  3

• 请问一下Iterative alignment是如何在代码中实现的呢？

  #23 opened 4 months ago by HeLeHanPrivate
  1

• divide and conquer

  #18 opened 5 months ago by dreaming12580
  1

• Performance of POPE?

  #17 opened 5 months ago by shufangxun
  1

• About optimizer setting in Iterative Alignment

  #19 opened 5 months ago by davidluciolu
  2

• 您好，我在跑训练的时候遇见了一个维度错误，不知道如何解决，我使用的是下载的数据集，clip是指定连接下载的clip

  #9 opened 5 months ago by XiaoLei2123
  3

• UnboundLocalError: local variable 'df' referenced before assignment

  #14 opened 5 months ago by Xuchen-Li
  2

• 请问下divide的模型是额外finetune过的么？

  #12 opened 6 months ago by menglin0320
  1

• Error when loading datasets split

  #13 opened 6 months ago by Xuchen-Li
  1

• control the output threshold of a large model

  #10 opened 6 months ago by zfr00
  1

• Hello,when I run the train-scripts,I met the dimension error.But

  #8 opened 6 months ago by XiaoLei2123
  0

• keyerror:idx,I have changed the data_dir,but when I run the train script,I occured the error.How to fix

  #7 opened 6 months ago by XiaoLei2123
  2

• Error loading the parquet dataset

  #5 opened 7 months ago by charismaticchiu
  3

• Is there a paper or tech report to show the details of training and dataset

  #2 opened 7 months ago by CrossLee1
  3

• Data file not normal when opening "obj_halbench_300_with_image.jsonl"

  #4 opened 7 months ago by hxhcreate
  2

• dpo_preference_processor not defined

  #3 opened 7 months ago by RifleZhang
  1

• ref_win_logp

  #1 opened 7 months ago by buptlihang
  2