File size: 7,273 Bytes
b33a043
 
 
 
 
80c4adf
b33a043
 
 
 
 
 
 
 
 
 
 
 
 
c9c9d19
 
 
 
 
 
 
 
 
b33a043
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c9c9d19
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b33a043
 
 
 
f021d10
b33a043
f021d10
 
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
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
---
library_name: transformers
tags:
- trl
- dpo
- alignment-handbook
- generated_from_trainer
model-index:
- name: OpenELM-1_1B-DPO-full-max-14-reward
  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. -->

# OpenELM-1_1B-DPO-full-max-14-reward

This model was trained from scratch on an unknown dataset.
It achieves the following results on the evaluation set:
- Loss: 1.1668
- Rewards/chosen: -3.5938
- Rewards/rejected: -4.0
- Rewards/accuracies: 0.4902
- Rewards/margins: 0.4121
- Logps/rejected: -688.0
- Logps/chosen: -676.0
- Logits/rejected: -16.375
- Logits/chosen: -16.875

## 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-05
- train_batch_size: 8
- eval_batch_size: 16
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- total_eval_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- 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 |
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
| 0.0562        | 0.1047 | 100  | 0.6971          | -1.2578        | -1.5703          | 0.5762             | 0.3145          | -446.0         | -444.0       | -9.3125         | -9.5625       |
| 0.0394        | 0.2094 | 200  | 0.7479          | -0.8516        | -1.0078          | 0.5195             | 0.1572          | -390.0         | -404.0       | -12.3125        | -12.75        |
| 0.0487        | 0.3141 | 300  | 0.9195          | -1.9922        | -2.3125          | 0.5176             | 0.3203          | -520.0         | -516.0       | -13.4375        | -13.6875      |
| 0.0454        | 0.4188 | 400  | 0.8309          | -1.4453        | -1.6016          | 0.4961             | 0.1543          | -448.0         | -462.0       | -15.625         | -15.75        |
| 0.0297        | 0.5236 | 500  | 0.8326          | -3.1094        | -3.375           | 0.5039             | 0.2734          | -628.0         | -628.0       | -15.5           | -15.6875      |
| 0.0434        | 0.6283 | 600  | 0.8373          | -1.6953        | -1.875           | 0.4941             | 0.1826          | -476.0         | -488.0       | -15.0           | -15.25        |
| 0.0496        | 0.7330 | 700  | 0.9407          | -3.7344        | -3.9688          | 0.5332             | 0.2236          | -684.0         | -692.0       | -9.5625         | -10.3125      |
| 0.0289        | 0.8377 | 800  | 1.0108          | -3.1406        | -3.25            | 0.4707             | 0.0991          | -612.0         | -632.0       | -13.0625        | -13.3125      |
| 0.0259        | 0.9424 | 900  | 1.0869          | -3.6094        | -3.7812          | 0.4648             | 0.1631          | -668.0         | -680.0       | -15.625         | -15.875       |
| 0.005         | 1.0471 | 1000 | 1.0944          | -3.4375        | -3.625           | 0.4570             | 0.1758          | -652.0         | -664.0       | -15.0625        | -15.25        |
| 0.0156        | 1.1518 | 1100 | 1.2452          | -4.4062        | -4.5938          | 0.4629             | 0.1973          | -748.0         | -760.0       | -16.5           | -16.625       |
| 0.0018        | 1.2565 | 1200 | 1.0496          | -3.7344        | -3.9219          | 0.4844             | 0.1885          | -680.0         | -692.0       | -15.5625        | -15.875       |
| 0.0046        | 1.3613 | 1300 | 1.0484          | -3.375         | -3.6094          | 0.4980             | 0.2402          | -648.0         | -656.0       | -14.9375        | -15.25        |
| 0.0041        | 1.4660 | 1400 | 0.9980          | -3.5156        | -3.8438          | 0.5137             | 0.3379          | -676.0         | -668.0       | -13.8125        | -14.3125      |
| 0.0077        | 1.5707 | 1500 | 1.0434          | -3.1719        | -3.5156          | 0.4902             | 0.3535          | -640.0         | -636.0       | -13.875         | -14.375       |
| 0.0016        | 1.6754 | 1600 | 1.0882          | -3.8594        | -4.2812          | 0.4922             | 0.4141          | -716.0         | -704.0       | -12.4375        | -12.9375      |
| 0.0042        | 1.7801 | 1700 | 1.0261          | -3.3438        | -3.7656          | 0.4941             | 0.4238          | -664.0         | -652.0       | -15.5           | -15.9375      |
| 0.0005        | 1.8848 | 1800 | 1.0536          | -3.2344        | -3.5938          | 0.4961             | 0.3555          | -648.0         | -644.0       | -16.625         | -17.0         |
| 0.0083        | 1.9895 | 1900 | 1.1039          | -3.4844        | -3.8125          | 0.4883             | 0.3242          | -672.0         | -668.0       | -16.25          | -16.625       |
| 0.0003        | 2.0942 | 2000 | 1.1159          | -3.5156        | -3.8438          | 0.4922             | 0.3301          | -672.0         | -672.0       | -16.125         | -16.625       |
| 0.0027        | 2.1990 | 2100 | 1.1535          | -3.5938        | -4.0             | 0.4980             | 0.4043          | -688.0         | -680.0       | -16.125         | -16.625       |
| 0.0003        | 2.3037 | 2200 | 1.1505          | -3.5781        | -3.9844          | 0.4902             | 0.4062          | -688.0         | -676.0       | -16.25          | -16.625       |
| 0.0006        | 2.4084 | 2300 | 1.1535          | -3.5469        | -3.9531          | 0.4902             | 0.4023          | -684.0         | -672.0       | -16.25          | -16.75        |
| 0.0002        | 2.5131 | 2400 | 1.1581          | -3.5781        | -3.9844          | 0.4922             | 0.4082          | -688.0         | -676.0       | -16.25          | -16.625       |
| 0.0001        | 2.6178 | 2500 | 1.1609          | -3.5625        | -3.9688          | 0.4961             | 0.4082          | -684.0         | -672.0       | -16.375         | -16.75        |
| 0.0008        | 2.7225 | 2600 | 1.1668          | -3.5938        | -4.0             | 0.4922             | 0.4121          | -688.0         | -676.0       | -16.375         | -16.75        |
| 0.0002        | 2.8272 | 2700 | 1.1668          | -3.5938        | -4.0             | 0.4902             | 0.4121          | -688.0         | -676.0       | -16.375         | -16.75        |
| 0.0003        | 2.9319 | 2800 | 1.1668          | -3.5938        | -4.0             | 0.4902             | 0.4121          | -688.0         | -676.0       | -16.375         | -16.875       |


### Framework versions

- Transformers 4.45.1
- Pytorch 2.3.0
- Datasets 3.0.1
- Tokenizers 0.20.0