YYYYYYibo's picture
DPO-7b-beta0.01
14d1055 verified
metadata
license: apache-2.0
library_name: peft
tags:
  - alignment-handbook
  - generated_from_trainer
  - trl
  - dpo
base_model: alignment-handbook/zephyr-7b-sft-full
datasets:
  - updated
  - original
model-index:
  - name: nash_dpo_rank4_on_vanilla_iter_3
    results: []

nash_dpo_rank4_on_vanilla_iter_3

This model is a fine-tuned version of YYYYYYibo/nash_dpo_rank4_on_vanilla_iter_2 on the updated and the original datasets. It achieves the following results on the evaluation set:

  • Loss: 0.6884
  • Rewards/chosen: -0.1839
  • Rewards/rejected: -0.2354
  • Rewards/accuracies: 0.5440
  • Rewards/margins: 0.0514
  • Logps/rejected: -311.2962
  • Logps/chosen: -324.9433
  • Logits/rejected: -2.0325
  • Logits/chosen: -2.1221

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 5e-06
  • train_batch_size: 2
  • eval_batch_size: 2
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 4
  • gradient_accumulation_steps: 16
  • total_train_batch_size: 128
  • total_eval_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 1

Training results

Training Loss Epoch Step Validation Loss Rewards/chosen Rewards/rejected Rewards/accuracies Rewards/margins Logps/rejected Logps/chosen Logits/rejected Logits/chosen
0.5749 0.61 100 0.6884 -0.1839 -0.2354 0.5440 0.0514 -311.2962 -324.9433 -2.0325 -2.1221

Framework versions

  • PEFT 0.7.1
  • Transformers 4.36.2
  • Pytorch 2.3.0+cu121
  • Datasets 2.14.6
  • Tokenizers 0.15.2