README.md

metadata

license: apache-2.0
base_model: mistralai/Mistral-7B-Instruct-v0.2
tags:
  - trl
  - dpo
  - generated_from_trainer
model-index:
  - name: >-
      DPO-PairRM-5-SMI-lr-1e6-iteration-5-t-7e-beta-15e3-1-iteration-baseline-D1-2e-D2_smi-1
    results: []

DPO-PairRM-5-SMI-lr-1e6-iteration-5-t-7e-beta-15e3-1-iteration-baseline-D1-2e-D2_smi-1

This model is a fine-tuned version of mistralai/Mistral-7B-Instruct-v0.2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7394
• Rewards/chosen: -2.1682
• Rewards/rejected: -2.3518
• Rewards/accuracies: 0.5835
• Rewards/margins: 0.1836
• Rewards/mix Margin: 0.0868
• Logps/rejected: -487.3550
• Logps/chosen: -484.3013
• Logits/rejected: -1.8268
• Logits/chosen: -1.8454

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: 1e-06
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 16
• total_train_batch_size: 64
• total_eval_batch_size: 4
• 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

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.2
• Datasets 2.17.1
• Tokenizers 0.15.1