---
base_model: princeton-nlp/Llama-3-Base-8B-SFT
library_name: peft
tags:
- alignment-handbook
- trl
- dpo
- generated_from_trainer
model-index:
- name: llama3-dpo-lora
  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. -->

# llama3-dpo-lora

This model is a fine-tuned version of [princeton-nlp/Llama-3-Base-8B-SFT](https://huggingface.co/princeton-nlp/Llama-3-Base-8B-SFT) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.5199
- Rewards/chosen: -0.1477
- Rewards/rejected: -0.9502
- Rewards/accuracies: 0.7260
- Rewards/margins: 0.8025
- Logps/rejected: -283.9596
- Logps/chosen: -291.2388
- Logits/rejected: -0.3914
- Logits/chosen: -0.4217

## 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-06
- train_batch_size: 1
- eval_batch_size: 4
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 16
- total_train_batch_size: 64
- total_eval_batch_size: 16
- 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 |
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
| 0.6297        | 0.1047 | 100  | 0.6140          | 0.1358         | -0.1277          | 0.6960             | 0.2634          | -275.7340      | -288.4034    | -0.5479         | -0.5526       |
| 0.5676        | 0.2094 | 200  | 0.5569          | -0.1144        | -0.6599          | 0.7000             | 0.5455          | -281.0560      | -290.9051    | -0.4945         | -0.5116       |
| 0.5414        | 0.3141 | 300  | 0.5403          | -0.3808        | -1.0461          | 0.7260             | 0.6652          | -284.9180      | -293.5698    | -0.4540         | -0.4775       |
| 0.5124        | 0.4187 | 400  | 0.5341          | -0.2337        | -0.9896          | 0.7040             | 0.7559          | -284.3532      | -292.0986    | -0.4243         | -0.4516       |
| 0.5529        | 0.5234 | 500  | 0.5260          | -0.2177        | -1.0037          | 0.7240             | 0.7861          | -284.4948      | -291.9380    | -0.3995         | -0.4290       |
| 0.53          | 0.6281 | 600  | 0.5244          | -0.0687        | -0.8583          | 0.7200             | 0.7895          | -283.0403      | -290.4489    | -0.4028         | -0.4317       |
| 0.5028        | 0.7328 | 700  | 0.5190          | -0.3357        | -1.1360          | 0.7320             | 0.8003          | -285.8177      | -293.1184    | -0.3874         | -0.4179       |
| 0.5347        | 0.8375 | 800  | 0.5191          | -0.1404        | -0.9419          | 0.7320             | 0.8015          | -283.8760      | -291.1650    | -0.3924         | -0.4225       |
| 0.4783        | 0.9422 | 900  | 0.5190          | -0.1399        | -0.9459          | 0.7260             | 0.8060          | -283.9163      | -291.1600    | -0.3917         | -0.4219       |


### Framework versions

- PEFT 0.7.1
- Transformers 4.44.2
- Pytorch 2.2.1+cu121
- Datasets 2.14.6
- Tokenizers 0.19.1