---
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-1-rpo
  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-1-rpo

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: 0.0240
- Rewards/chosen: -0.1305
- Rewards/rejected: -0.3804
- Rewards/accuracies: 0.7328
- Rewards/margins: 0.2499
- Logps/rejected: -286.3204
- Logps/chosen: -273.0143
- Logits/rejected: -2.4857
- Logits/chosen: -2.5477

## 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 |
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
| 0.0371        | 0.1143 | 50   | 0.0330          | 0.0162         | -0.0844          | 0.6638             | 0.1006          | -256.7177      | -258.3393    | -2.5728         | -2.6169       |
| 0.0326        | 0.2286 | 100  | 0.0284          | -0.1604        | -0.3762          | 0.7069             | 0.2158          | -285.9020      | -276.0061    | -2.2908         | -2.3469       |
| 0.0276        | 0.3429 | 150  | 0.0261          | -0.1426        | -0.3607          | 0.7198             | 0.2181          | -284.3463      | -274.2200    | -2.3538         | -2.3992       |
| 0.0257        | 0.4571 | 200  | 0.0255          | -0.1250        | -0.3457          | 0.7371             | 0.2206          | -282.8442      | -272.4646    | -2.4669         | -2.5054       |
| 0.0259        | 0.5714 | 250  | 0.0249          | -0.1421        | -0.3761          | 0.7457             | 0.2340          | -285.8867      | -274.1743    | -2.5651         | -2.6093       |
| 0.0244        | 0.6857 | 300  | 0.0244          | -0.1115        | -0.3601          | 0.7328             | 0.2486          | -284.2852      | -271.1066    | -2.4931         | -2.5556       |
| 0.0232        | 0.8    | 350  | 0.0241          | -0.1203        | -0.3689          | 0.7328             | 0.2486          | -285.1674      | -271.9947    | -2.4940         | -2.5542       |
| 0.0253        | 0.9143 | 400  | 0.0240          | -0.1305        | -0.3804          | 0.7328             | 0.2499          | -286.3204      | -273.0143    | -2.4857         | -2.5477       |


### Framework versions

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