princeton-nlp/SimPO

Issues

• How to run the evaluation on GPT-4 Turbo

  #49 opened 6 months ago by Xalp
  2

• looking for the model parameters of Llama3-Instruct (8B) SFT

  #63 opened 4 months ago by alphatogo
  1

• Difference with changing the gradient accumulation - ZeroEval and AlpacaEval 2

  #61 opened 5 months ago by sahsaeedi
  4

• Question about the `annotators_config` and `reference_outputs` in alpaca_eval

  #55 opened 5 months ago by AIR-hl
  1

• On-Policy Preference Data Generation

  #62 opened 5 months ago by tannedbum
  2

• Evaluate about arena-hard.

  #57 opened 5 months ago by hitszxs
  3

• reward/chosen is decreasing

  #42 opened 6 months ago by zhangguoxin1
  6

• Lora about simpo.

  #39 opened 6 months ago by hitszxs
  6

• Unable to reproduce the results of SFT

  #27 opened 6 months ago by yujiaw98
  8

• Where is length normalization in the code?

  #59 opened 5 months ago by ElegantLin
  2

• Cannot reproduce the training curves of LLaMA3-simpo-v2

  #60 opened 5 months ago by jdf-prog
  2

• Can you share the loss log of `Llama-3-8B-Instruct`

  #45 opened 6 months ago by AIR-hl
  3

• ArgumentError: `attn_implementation`

  #58 opened 5 months ago by ElegantLin
  2

• What version of Mistral 7B was used for princeton-nlp/Mistral-7B-Instruct-SimPO ?

  #56 opened 5 months ago by junkangwu
  1

• can't reproduce the results of Mistral-7B-Instruct-DPO

  #38 opened 6 months ago by RikkiXu
  10

• bug using accelerate

  #54 opened 6 months ago by cjakfskvnad
  0

• Yi-34B + Simpo + Full + Novel writing Task, fine-tuning results are unsatisfactory

  #33 opened 6 months ago by onlyfish79
  3

• Why is the AlpacaEval2 score of meta-llama/Meta-LLama-3-8B-Instruct in the paper higher than that on the leaderboard?

  #31 opened 6 months ago by eugene-yh
  1

• DPOTrainer.get_batch_logps() got an unexpected keyword argument 'average_log_prob'

  #29 opened 6 months ago by RAY2L
  2

• Unable to reproduce the mt bench results in the paper

  #22 opened 6 months ago by lmx760581375
  1

• Has anyone reproduced the results in the paper quantitatively?

  #21 opened 6 months ago by binzhwang
  7

• For the Instruct setup, why do different models require different training datasets? Can the same dataset be used?

  #19 opened 6 months ago by qiuwenbogdut
  3

• the outputs of reproduced model has "<|start_header_id|>assistant<|end_header_id|>" at beginning

  #17 opened 6 months ago by binzhwang
  6

• 'loss': 0.0, 'grad_norm': nan, and get

  #35 opened 6 months ago by wujia11
  1

• Hi, could you please provide the exact versions of the packages needed to reproduce the results and your released checkpoint?

  #40 opened 6 months ago by AGTSAAA
  1

• Use a better `ultrafeedback`

  #43 opened 6 months ago by AIR-hl
  1

• Questions about recent changes

  #51 opened 6 months ago by junkangwu
  2

• Model Request: Meta-Llama-3.1-8B-Instruct

  #52 opened 6 months ago by Bearsaerker
  1

• No template of gemma2 when evaluating it on AE2

  #53 opened 6 months ago by sunjie279
  1

• Could you offer the packages' version of alpaca_eval 2 and the generated answers?

  #44 opened 6 months ago by sunjie279
  1

• Cannot install alignment-handbook==0.4.0

  #48 opened 6 months ago by Shentao-YANG
  1

• Any technique for hyperparameter tuning?

  #50 opened 6 months ago by NitinAB1108
  2

• Could you add licenses to the preference datasets after reward model labeling on huggingface?

  #46 opened 6 months ago by hanyang1999
  3

• Experimental results on ARC-C subset for challeging reasoning?

  #47 opened 6 months ago by tongyx361
  0

• How to use local dataset

  #41 opened 6 months ago by mazhengyufreedom
  3

• About label_smoothing

  #36 opened 6 months ago by mazhengyufreedom
  2

• AttributeError: 'SimPOConfig' object has no attribute 'ref_model_init_kwargs'. Did you mean 'model_init_kwargs'?

  #34 opened 6 months ago by Saumajit
  2

• Why do the same methods have different results?

  #37 opened 6 months ago by XiepengLi
  2

• Unable to reproduce the results of DPO

  #23 opened 6 months ago by AGTSAAA
  4

• Question about apply_chat_template to prompt

  #30 opened 6 months ago by EganGu
  2

• Your comparsion is unfair by using different chat templates

  #24 opened 6 months ago by AGTSAAA
  2

• Usage on Custom Dataset

  #28 opened 6 months ago by ViperVille007
  1

• QLORA -4bit

  #14 opened 6 months ago by seroetr
  1

• Question about the length-normalization

  #13 opened 6 months ago by yujiaw98
  1

• simloss has no length normalization in simpo loss

  #32 opened 6 months ago by hjc3613
  3

• Change of Mistral Chat template

  #26 opened 7 months ago by mianzhang
  2

• TRL version should be 0.8.6

  #25 opened 7 months ago by blakechi
  2

• Length normalization in DPO and other variants

  #20 opened 7 months ago by yakazimir
  2

• ValueError: Unknown split "train". Should be one of ['train_iteration_1', 'test_iteration_1', 'train_iteration_2', 'test_iteration_2',

  #18 opened 7 months ago by qiuwenbogdut
  2

• Confusing Code logic

  #15 opened 7 months ago by syboomsy
  2