---
license: apache-2.0
base_model: tsavage68/UTI_M2_1000steps_1e5rate_SFT
tags:
- trl
- dpo
- generated_from_trainer
model-index:
- name: UTI_M2_200steps_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. -->

# UTI_M2_200steps_1e7rate_01beta_CSFTDPO

This model is a fine-tuned version of [tsavage68/UTI_M2_1000steps_1e5rate_SFT](https://huggingface.co/tsavage68/UTI_M2_1000steps_1e5rate_SFT) on an unknown dataset.
It achieves the following results on the evaluation set:
- Loss: 0.2079
- Rewards/chosen: 0.4687
- Rewards/rejected: -1.5148
- Rewards/accuracies: 0.9000
- Rewards/margins: 1.9835
- Logps/rejected: -59.3138
- Logps/chosen: -15.6076
- Logits/rejected: -3.8213
- Logits/chosen: -3.7441

## 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: 200

### Training results

| Training Loss | Epoch  | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
| 0.6928        | 0.3333 | 25   | 0.6881          | 0.0023         | -0.0079          | 0.7300             | 0.0102          | -44.2452       | -20.2718     | -3.8169         | -3.7449       |
| 0.6578        | 0.6667 | 50   | 0.6325          | 0.0226         | -0.1063          | 0.8800             | 0.1289          | -45.2294       | -20.0689     | -3.8170         | -3.7449       |
| 0.5691        | 1.0    | 75   | 0.5043          | 0.0873         | -0.3772          | 0.9000             | 0.4645          | -47.9383       | -19.4217     | -3.8188         | -3.7462       |
| 0.3598        | 1.3333 | 100  | 0.3707          | 0.1758         | -0.7982          | 0.8900             | 0.9740          | -52.1479       | -18.5365     | -3.8221         | -3.7487       |
| 0.2615        | 1.6667 | 125  | 0.2615          | 0.3619         | -1.2250          | 0.9000             | 1.5869          | -56.4161       | -16.6756     | -3.8216         | -3.7458       |
| 0.2369        | 2.0    | 150  | 0.2166          | 0.4540         | -1.4578          | 0.9000             | 1.9118          | -58.7439       | -15.7542     | -3.8215         | -3.7446       |
| 0.1778        | 2.3333 | 175  | 0.2082          | 0.4687         | -1.5135          | 0.9000             | 1.9822          | -59.3007       | -15.6071     | -3.8212         | -3.7441       |
| 0.185         | 2.6667 | 200  | 0.2079          | 0.4687         | -1.5148          | 0.9000             | 1.9835          | -59.3138       | -15.6076     | -3.8213         | -3.7441       |


### Framework versions

- Transformers 4.41.2
- Pytorch 2.0.0+cu117
- Datasets 2.19.2
- Tokenizers 0.19.1