File size: 2,958 Bytes
4190f79
 
 
 
 
10dc65c
4190f79
 
 
10dc65c
 
 
 
 
4190f79
 
 
 
 
 
 
 
 
 
10dc65c
4190f79
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
---
library_name: transformers
license: gemma
base_model: google/gemma-7b
tags:
- alignment-handbook
- trl
- orpo
- generated_from_trainer
- trl
- orpo
- generated_from_trainer
datasets:
- silviasapora/low_quality_dpo7k
model-index:
- name: gemma-7b-borpo-low-quality-v2
  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. -->

# gemma-7b-borpo-low-quality-v2

This model is a fine-tuned version of [google/gemma-7b](https://huggingface.co/google/gemma-7b) on the silviasapora/low_quality_dpo7k dataset.
It achieves the following results on the evaluation set:
- Loss: 1.6017
- Rewards/chosen: -0.0578
- Rewards/rejected: -0.0690
- Rewards/accuracies: 0.5714
- Rewards/margins: 0.0112
- Logps/rejected: -1.3795
- Logps/chosen: -1.1561
- Logits/rejected: 249.0934
- Logits/chosen: 304.2649
- Nll Loss: 1.5643
- Log Odds Ratio: -0.6745
- Log Odds Chosen: 0.3316

## 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: 2
- eval_batch_size: 1
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- total_eval_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: inverse_sqrt
- lr_scheduler_warmup_steps: 100
- num_epochs: 3

### Training results

| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen | Nll Loss | Log Odds Ratio | Log Odds Chosen |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|:--------:|:--------------:|:---------------:|
| 1.4218        | 1.0   | 168  | 1.4488          | -0.0504        | -0.0580          | 0.5571             | 0.0076          | -1.1591        | -1.0071      | 273.7526        | 326.8029      | 1.4553   | -0.6712        | 0.2324          |
| 1.0804        | 2.0   | 336  | 1.4225          | -0.0511        | -0.0591          | 0.5143             | 0.0080          | -1.1830        | -1.0220      | 278.2473        | 330.5067      | 1.4083   | -0.6897        | 0.2152          |
| 0.5651        | 3.0   | 504  | 1.6017          | -0.0578        | -0.0690          | 0.5714             | 0.0112          | -1.3795        | -1.1561      | 249.0934        | 304.2649      | 1.5643   | -0.6745        | 0.3316          |


### Framework versions

- Transformers 4.44.2
- Pytorch 2.4.1+cu121
- Datasets 3.0.0
- Tokenizers 0.19.1