---
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 an unknown dataset.
It achieves the following results on the evaluation set:
- Loss: 0.5152
- Rewards/chosen: -1.1647
- Rewards/rejected: -2.0467
- Rewards/accuracies: 0.7599
- Rewards/margins: 0.8821
- Logps/rejected: -470.8255
- Logps/chosen: -408.1943
- Logits/rejected: 1.9026
- Logits/chosen: 1.3480

## 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: 4
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- total_eval_batch_size: 32
- 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 |
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
| 0.6509        | 0.1047 | 100  | 0.6507          | -0.0317        | -0.1403          | 0.6865             | 0.1086          | -280.1782      | -294.8975    | -2.3950         | -2.4570       |
| 0.5842        | 0.2093 | 200  | 0.5869          | -0.5628        | -1.0165          | 0.7183             | 0.4537          | -367.8032      | -348.0123    | -1.8883         | -1.9913       |
| 0.5518        | 0.3140 | 300  | 0.5573          | -0.7616        | -1.3682          | 0.7421             | 0.6066          | -402.9727      | -367.8922    | -1.4401         | -1.5902       |
| 0.5596        | 0.4186 | 400  | 0.5451          | -0.5483        | -1.1855          | 0.7579             | 0.6373          | -384.7072      | -346.5541    | -1.3655         | -1.5387       |
| 0.5352        | 0.5233 | 500  | 0.5309          | -0.9240        | -1.6733          | 0.7440             | 0.7493          | -433.4807      | -384.1307    | -0.0937         | -0.4008       |
| 0.4998        | 0.6279 | 600  | 0.5225          | -1.0419        | -1.8718          | 0.7401             | 0.8299          | -453.3297      | -395.9146    | 1.1206          | 0.6984        |
| 0.4936        | 0.7326 | 700  | 0.5196          | -1.2088        | -2.0934          | 0.7540             | 0.8846          | -475.4923      | -412.6063    | 1.9631          | 1.4184        |
| 0.5125        | 0.8373 | 800  | 0.5171          | -1.1926        | -2.0718          | 0.7659             | 0.8791          | -473.3307      | -410.9924    | 1.8737          | 1.3314        |
| 0.4976        | 0.9419 | 900  | 0.5152          | -1.1647        | -2.0467          | 0.7599             | 0.8821          | -470.8255      | -408.1943    | 1.9026          | 1.3480        |


### Framework versions

- Transformers 4.40.2
- Pytorch 2.1.2
- Datasets 2.19.1
- Tokenizers 0.19.1