gemma-dpo / README.md
redscroll's picture
Model save
b1856bb verified
metadata
license: apache-2.0
library_name: peft
tags:
  - trl
  - dpo
  - generated_from_trainer
base_model: cognitivecomputations/dolphin-2.1-mistral-7b
model-index:
  - name: gemma-dpo
    results: []

gemma-dpo

This model is a fine-tuned version of cognitivecomputations/dolphin-2.1-mistral-7b on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.6646
  • Rewards/chosen: -9.3932
  • Rewards/rejected: -13.4385
  • Rewards/accuracies: 0.7852
  • Rewards/margins: 4.0453
  • Logps/rejected: -440.0019
  • Logps/chosen: -450.6083
  • Logits/rejected: -2.5347
  • Logits/chosen: -2.6367

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-05
  • train_batch_size: 12
  • eval_batch_size: 4
  • seed: 42
  • 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
0.7403 0.76 700 0.6646 -9.3932 -13.4385 0.7852 4.0453 -440.0019 -450.6083 -2.5347 -2.6367

Framework versions

  • PEFT 0.9.0
  • Transformers 4.38.2
  • Pytorch 2.2.1+cu121
  • Datasets 2.18.0
  • Tokenizers 0.15.2