---
license: apache-2.0
base_model: alignment-handbook/zephyr-7b-sft-full
tags:
- trl
- dpo
- generated_from_trainer
model-index:
- name: zephyr-7b-dpo-full-prometheus-reward-scale-05
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# zephyr-7b-dpo-full-prometheus-reward-scale-05

This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.5142
- Rewards/chosen: -1.9575
- Rewards/rejected: -3.2488
- Rewards/accuracies: 0.7198
- Rewards/margins: 1.2913
- Logps/rejected: -573.1553
- Logps/chosen: -455.7142
- Logits/rejected: 4.0394
- Logits/chosen: 3.0591

## 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-07
- train_batch_size: 8
- eval_batch_size: 8
- seed: 55
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 2
- total_train_batch_size: 128
- total_eval_batch_size: 64
- 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 |
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
| 0.6718        | 0.1143 | 50   | 0.6623          | -0.0203        | -0.1341          | 0.6595             | 0.1139          | -261.6897      | -261.9885    | -2.5701         | -2.6137       |
| 0.5837        | 0.2286 | 100  | 0.5855          | -0.7924        | -1.5143          | 0.6595             | 0.7219          | -399.7070      | -339.2018    | -0.2076         | -0.5772       |
| 0.5481        | 0.3429 | 150  | 0.5498          | -1.3440        | -2.2915          | 0.6940             | 0.9475          | -477.4253      | -394.3634    | 2.4336          | 1.6485        |
| 0.5289        | 0.4571 | 200  | 0.5409          | -1.6426        | -2.7545          | 0.6983             | 1.1119          | -523.7253      | -424.2230    | 3.5667          | 2.6372        |
| 0.54          | 0.5714 | 250  | 0.5280          | -1.5391        | -2.7058          | 0.7026             | 1.1667          | -518.8563      | -413.8667    | 2.6437          | 1.4166        |
| 0.5147        | 0.6857 | 300  | 0.5204          | -1.8487        | -3.0990          | 0.7112             | 1.2504          | -558.1808      | -444.8300    | 3.7771          | 2.6969        |
| 0.5033        | 0.8    | 350  | 0.5168          | -1.8756        | -3.1461          | 0.7414             | 1.2705          | -562.8871      | -447.5259    | 3.8637          | 2.7948        |
| 0.52          | 0.9143 | 400  | 0.5142          | -1.9575        | -3.2488          | 0.7198             | 1.2913          | -573.1553      | -455.7142    | 4.0394          | 3.0591        |


### Framework versions

- Transformers 4.44.0.dev0
- Pytorch 2.1.2
- Datasets 2.20.0
- Tokenizers 0.19.1