---
library_name: transformers
license: llama3
base_model: tsavage68/Na_L3_100steps_1e6rate_SFT
tags:
- trl
- dpo
- generated_from_trainer
model-index:
- name: Na_L3_350steps_1e7rate_01beta_cSFTDPO
  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. -->

# Na_L3_350steps_1e7rate_01beta_cSFTDPO

This model is a fine-tuned version of [tsavage68/Na_L3_100steps_1e6rate_SFT](https://huggingface.co/tsavage68/Na_L3_100steps_1e6rate_SFT) on an unknown dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0088
- Rewards/chosen: 0.6901
- Rewards/rejected: -4.1623
- Rewards/accuracies: 1.0
- Rewards/margins: 4.8524
- Logps/rejected: -83.1246
- Logps/chosen: -17.9889
- Logits/rejected: -0.9502
- Logits/chosen: -0.8819

## 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: 1e-07
- train_batch_size: 2
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 100
- training_steps: 350

### Training results

| Training Loss | Epoch  | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
| 0.6442        | 0.2667 | 50   | 0.6172          | 0.0288         | -0.1304          | 1.0                | 0.1591          | -42.8057       | -24.6026     | -0.9518         | -0.8851       |
| 0.2897        | 0.5333 | 100  | 0.2504          | 0.1177         | -1.1534          | 1.0                | 1.2711          | -53.0359       | -23.7135     | -0.9534         | -0.8871       |
| 0.0587        | 0.8    | 150  | 0.0469          | 0.4687         | -2.7071          | 1.0                | 3.1758          | -68.5731       | -20.2031     | -0.9553         | -0.8874       |
| 0.0185        | 1.0667 | 200  | 0.0155          | 0.6102         | -3.6824          | 1.0                | 4.2926          | -78.3254       | -18.7883     | -0.9531         | -0.8845       |
| 0.0097        | 1.3333 | 250  | 0.0096          | 0.6743         | -4.0935          | 1.0                | 4.7678          | -82.4367       | -18.1468     | -0.9518         | -0.8835       |
| 0.0083        | 1.6    | 300  | 0.0088          | 0.6862         | -4.1645          | 1.0                | 4.8507          | -83.1466       | -18.0285     | -0.9504         | -0.8819       |
| 0.0079        | 1.8667 | 350  | 0.0088          | 0.6901         | -4.1623          | 1.0                | 4.8524          | -83.1246       | -17.9889     | -0.9502         | -0.8819       |


### Framework versions

- Transformers 4.44.2
- Pytorch 2.4.0+cu121
- Datasets 2.21.0
- Tokenizers 0.19.1