---
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: 0.5090
- Rewards/chosen: -1.1007
- Rewards/rejected: -2.0002
- Rewards/accuracies: 0.7738
- Rewards/margins: 0.8995
- Logps/rejected: -466.1724
- Logps/chosen: -401.8018
- Logits/rejected: 3.6229
- Logits/chosen: 2.8669

## 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 |
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
| 0.6512        | 0.1047 | 100  | 0.6511          | -0.0190        | -0.1266          | 0.6964             | 0.1076          | -278.8090      | -293.6257    | -2.3851         | -2.4490       |
| 0.5992        | 0.2093 | 200  | 0.5944          | -0.2668        | -0.6535          | 0.7103             | 0.3866          | -331.5005      | -318.4129    | -1.7454         | -1.8605       |
| 0.5469        | 0.3140 | 300  | 0.5530          | -0.6557        | -1.3199          | 0.7520             | 0.6642          | -398.1460      | -357.2993    | -0.7401         | -0.9693       |
| 0.5491        | 0.4186 | 400  | 0.5448          | -1.0399        | -1.6860          | 0.7282             | 0.6462          | -434.7570      | -395.7156    | 1.3254          | 0.9052        |
| 0.5351        | 0.5233 | 500  | 0.5296          | -0.8199        | -1.6144          | 0.7679             | 0.7945          | -427.5919      | -373.7142    | 2.7946          | 2.2107        |
| 0.4879        | 0.6279 | 600  | 0.5152          | -1.1813        | -2.0574          | 0.7619             | 0.8761          | -471.8891      | -409.8589    | 3.3049          | 2.6265        |
| 0.4963        | 0.7326 | 700  | 0.5121          | -1.1447        | -2.0602          | 0.7679             | 0.9156          | -472.1772      | -406.1937    | 3.7355          | 2.9642        |
| 0.5009        | 0.8373 | 800  | 0.5099          | -1.1326        | -2.0244          | 0.7679             | 0.8919          | -468.5970      | -404.9855    | 3.6202          | 2.8807        |
| 0.4926        | 0.9419 | 900  | 0.5090          | -1.1007        | -2.0002          | 0.7738             | 0.8995          | -466.1724      | -401.8018    | 3.6229          | 2.8669        |


### Framework versions

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