---
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: 1.4971
- Rewards/chosen: -4.5102
- Rewards/rejected: -4.6591
- Rewards/accuracies: 0.5156
- Rewards/margins: 0.1490
- Logps/rejected: -753.6738
- Logps/chosen: -732.6489
- Logits/rejected: 1.5926
- Logits/chosen: 1.5057

## 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 | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
| 0.2876        | 0.1   | 100  | -2.3965       | -2.3559         | -391.6134    | -394.6287      | 0.8317          | 0.4883             | -1.0998        | -0.0311         | -1.0687          |
| 0.1728        | 0.21  | 200  | -0.2344       | -0.1269         | -464.6779    | -471.8403      | 1.0232          | 0.4766             | -1.8304        | 0.0103          | -1.8408          |
| 0.1485        | 0.31  | 300  | -0.3320       | -0.2139         | -506.0840    | -508.1475      | 1.1085          | 0.4883             | -2.2445        | -0.0406         | -2.2039          |
| 0.1363        | 0.42  | 400  | -0.2901       | -0.1728         | -477.3530    | -486.5422      | 1.1616          | 0.4961             | -1.9572        | 0.0306          | -1.9878          |
| 0.1192        | 0.52  | 500  | 0.8077        | 0.8821          | -553.1240    | -562.3370      | 1.2602          | 0.4961             | -2.7149        | 0.0308          | -2.7458          |
| 0.1061        | 0.63  | 600  | 1.3570        | -3.5510         | -3.6801      | 0.5078         | 0.1291          | -655.7740          | -636.7335      | 1.4624          | 1.3499           |
| 0.0916        | 0.73  | 700  | 1.5923        | -4.9928         | -5.1535      | 0.5195         | 0.1607          | -803.1144          | -780.9122      | 1.8370          | 1.7244           |
| 0.0982        | 0.84  | 800  | 1.4367        | -4.1982         | -4.3446      | 0.5117         | 0.1464          | -722.2228          | -701.4560      | 1.5885          | 1.4960           |
| 0.0798        | 0.94  | 900  | 1.4971        | -4.5102         | -4.6591      | 0.5156         | 0.1490          | -753.6738          | -732.6489      | 1.5926          | 1.5057           |


### Framework versions

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