---
tags:
- trl
- dpo
- generated_from_trainer
model-index:
- name: zephyr-7b-dpo-full
  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

This model was trained from scratch on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 2.0232
- Rewards/chosen: -7.1946
- Rewards/rejected: -8.7238
- Rewards/accuracies: 0.6133
- Rewards/margins: 1.5292
- Logps/rejected: -1160.1461
- Logps/chosen: -1001.0963
- Logits/rejected: -0.4190
- Logits/chosen: -0.5892

## 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: 42
- 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 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
| 0.254         | 0.1   | 100  | 1.4761          | -5.3290        | -6.2112          | 0.5898             | 0.8822          | -908.8818      | -814.5385    | -1.4873         | -1.5203       |
| 0.1844        | 0.21  | 200  | 1.7253          | -6.1555        | -7.4481          | 0.6133             | 1.2926          | -1032.5726     | -897.1824    | -1.4103         | -1.4664       |
| 0.1635        | 0.31  | 300  | 1.6677          | -6.1768        | -7.3921          | 0.5938             | 1.2153          | -1026.9750     | -899.3143    | -0.6257         | -0.7515       |
| 0.1606        | 0.42  | 400  | 2.0307          | -7.0774        | -8.4601          | 0.6016             | 1.3827          | -1133.7700     | -989.3718    | -0.4798         | -0.6143       |
| 0.163         | 0.52  | 500  | 1.8216          | -6.5495        | -8.0368          | 0.5898             | 1.4873          | -1091.4379     | -936.5793    | -0.8136         | -0.9380       |
| 0.1656        | 0.63  | 600  | 1.8091          | -6.5309        | -7.9593          | 0.625              | 1.4284          | -1083.6920     | -934.7285    | -0.3700         | -0.5360       |
| 0.1552        | 0.73  | 700  | 2.0767          | -7.6318        | -9.1866          | 0.5977             | 1.5547          | -1206.4197     | -1044.8179   | -0.3588         | -0.5351       |
| 0.1377        | 0.84  | 800  | 2.0307          | -7.2870        | -8.8043          | 0.6055             | 1.5173          | -1168.1901     | -1010.3356   | -0.4156         | -0.5887       |
| 0.1462        | 0.94  | 900  | 2.0232          | -7.1946        | -8.7238          | 0.6133             | 1.5292          | -1160.1461     | -1001.0963   | -0.4190         | -0.5892       |


### Framework versions

- Transformers 4.38.2
- Pytorch 2.1.2+cu118
- Datasets 2.16.1
- Tokenizers 0.15.2