---
base_model: HuggingFaceH4/zephyr-7b-gemma-sft-v0.1
datasets:
- argilla/dpo-mix-7k
- RedaAlami/PKU-SafeRLHF-Processed
library_name: peft
license: other
tags:
- alignment-handbook
- trl
- dpo
- generated_from_trainer
model-index:
- name: zephyr-7b-gemma-dpo
  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-gemma-dpo

This model is a fine-tuned version of [HuggingFaceH4/zephyr-7b-gemma-sft-v0.1](https://huggingface.co/HuggingFaceH4/zephyr-7b-gemma-sft-v0.1) on the argilla/dpo-mix-7k and the RedaAlami/PKU-SafeRLHF-Processed datasets.
It achieves the following results on the evaluation set:
- Loss: 0.6478
- Rewards/chosen: -0.3452
- Rewards/rejected: -0.5788
- Rewards/accuracies: 0.6169
- Rewards/margins: 0.2336
- Logps/rejected: -334.5554
- Logps/chosen: -295.9647
- Logits/rejected: 436.0139
- Logits/chosen: 452.6414

## 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: 2
- eval_batch_size: 4
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 8
- total_train_batch_size: 128
- 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: 2

### Training results

| Training Loss | Epoch  | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
| 0.7052        | 0.2602 | 100  | 0.7032          | 0.1828         | 0.1374           | 0.5148             | 0.0454          | -320.2303      | -285.4035    | 437.8931        | 454.4476      |
| 0.6851        | 0.5205 | 200  | 0.6794          | 0.1534         | 0.0240           | 0.5991             | 0.1294          | -322.4987      | -285.9917    | 436.5674        | 453.2177      |
| 0.6545        | 0.7807 | 300  | 0.6632          | -0.0335        | -0.2290          | 0.5962             | 0.1955          | -327.5587      | -289.7299    | 435.9517        | 452.6035      |
| 0.6428        | 1.0410 | 400  | 0.6532          | -0.3583        | -0.5844          | 0.6154             | 0.2261          | -334.6671      | -296.2265    | 436.0371        | 452.6768      |
| 0.6366        | 1.3012 | 500  | 0.6521          | -0.3063        | -0.5602          | 0.6124             | 0.2539          | -334.1831      | -295.1856    | 436.1843        | 452.8112      |
| 0.6058        | 1.5615 | 600  | 0.6497          | -0.3389        | -0.5751          | 0.6139             | 0.2362          | -334.4804      | -295.8380    | 436.0276        | 452.6521      |
| 0.6368        | 1.8217 | 700  | 0.6449          | -0.3403        | -0.5854          | 0.6065             | 0.2451          | -334.6864      | -295.8665    | 436.0117        | 452.6395      |


### Framework versions

- PEFT 0.12.0
- Transformers 4.43.3
- Pytorch 2.4.0+cu121
- Datasets 2.20.0
- Tokenizers 0.19.1