File size: 7,239 Bytes
60b4bb5
 
 
 
 
82d7cd6
60b4bb5
 
 
 
 
 
 
 
 
 
 
 
 
82d7cd6
 
 
 
60b4bb5
 
82d7cd6
60b4bb5
82d7cd6
60b4bb5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
82d7cd6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
60b4bb5
 
 
 
 
 
 
 
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-SLiC
  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-SLiC

This model was trained from scratch on an unknown dataset.
It achieves the following results on the evaluation set:
- Logits/chosen: -10.0625
- Logits/rejected: -8.75
- Logps/chosen: -752.0
- Logps/rejected: -824.0
- Loss: 0.6883
- Rewards/accuracies: 0.7344
- Rewards/chosen: -4.3438
- Rewards/margins: 0.9922
- Rewards/rejected: -5.3438

## 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 | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
|:-------------:|:------:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
| 0.7634        | 0.1047 | 100  | -13.0625      | -12.9375        | -392.0       | -392.0         | 0.7878          | 0.6406             | -0.7461        | 0.2832          | -1.0312          |
| 0.7498        | 0.2093 | 200  | -12.75        | -12.4375        | -436.0       | -444.0         | 0.7468          | 0.6719             | -1.1719        | 0.3809          | -1.5547          |
| 0.8142        | 0.3140 | 300  | -14.8125      | -14.75          | -504.0       | -516.0         | 0.7466          | 0.6914             | -1.8594        | 0.4141          | -2.2812          |
| 0.7764        | 0.4186 | 400  | -14.5625      | -14.4375        | -516.0       | -528.0         | 0.7499          | 0.6699             | -1.9688        | 0.4316          | -2.4062          |
| 0.731         | 0.5233 | 500  | -11.0         | -10.5           | -560.0       | -576.0         | 0.7240          | 0.6914             | -2.4219        | 0.4375          | -2.8594          |
| 0.665         | 0.6279 | 600  | -10.75        | -10.0625        | -660.0       | -696.0         | 0.7045          | 0.6973             | -3.4062        | 0.6680          | -4.0625          |
| 0.6806        | 0.7326 | 700  | -13.875       | -13.4375        | -568.0       | -604.0         | 0.6912          | 0.7070             | -2.5156        | 0.6523          | -3.1562          |
| 0.6597        | 0.8373 | 800  | -13.5         | -13.3125        | -548.0       | -576.0         | 0.7087          | 0.6777             | -2.2969        | 0.5664          | -2.8594          |
| 0.7325        | 0.9419 | 900  | -14.0         | -13.25          | -588.0       | -624.0         | 0.6838          | 0.7090             | -2.6875        | 0.6602          | -3.3594          |
| 0.2677        | 1.0466 | 1000 | -12.1875      | -11.0625        | -640.0       | -688.0         | 0.6726          | 0.7070             | -3.2344        | 0.7734          | -4.0             |
| 0.2256        | 1.1512 | 1100 | -11.125       | -10.0625        | -676.0       | -728.0         | 0.6992          | 0.7090             | -3.5938        | 0.7969          | -4.375           |
| 0.1954        | 1.2559 | 1200 | -11.3125      | -10.125         | -664.0       | -720.0         | 0.7033          | 0.7051             | -3.4688        | 0.8477          | -4.3125          |
| 0.2289        | 1.3605 | 1300 | -11.0         | -9.9375         | -692.0       | -740.0         | 0.6722          | 0.7344             | -3.7344        | 0.7852          | -4.5             |
| 0.2227        | 1.4652 | 1400 | -12.5         | -11.8125        | -676.0       | -720.0         | 0.6925          | 0.6953             | -3.5781        | 0.7383          | -4.3125          |
| 0.1902        | 1.5699 | 1500 | -12.0625      | -11.125         | -736.0       | -792.0         | 0.6758          | 0.7148             | -4.1875        | 0.8320          | -5.0312          |
| 0.2192        | 1.6745 | 1600 | -13.625       | -12.875         | -704.0       | -748.0         | 0.6833          | 0.7148             | -3.8438        | 0.7695          | -4.625           |
| 0.2137        | 1.7792 | 1700 | -11.9375      | -11.0           | -716.0       | -764.0         | 0.6734          | 0.7207             | -3.9688        | 0.8008          | -4.7812          |
| 0.2001        | 1.8838 | 1800 | -12.125       | -11.3125        | -692.0       | -740.0         | 0.6734          | 0.7207             | -3.7344        | 0.7617          | -4.5             |
| 0.1713        | 1.9885 | 1900 | -10.4375      | -9.25           | -712.0       | -768.0         | 0.6680          | 0.7383             | -3.9375        | 0.8789          | -4.8125          |
| 0.0184        | 2.0931 | 2000 | -11.0625      | -9.875          | -704.0       | -768.0         | 0.6845          | 0.7305             | -3.8594        | 0.9453          | -4.8125          |
| 0.0313        | 2.1978 | 2100 | -11.25        | -10.125         | -720.0       | -784.0         | 0.6798          | 0.7402             | -4.0           | 0.9570          | -4.9688          |
| 0.0401        | 2.3025 | 2200 | -10.6875      | -9.375          | -732.0       | -800.0         | 0.6865          | 0.7363             | -4.1562        | 0.9492          | -5.0938          |
| 0.0211        | 2.4071 | 2300 | -10.125       | -8.75           | -740.0       | -812.0         | 0.6874          | 0.7383             | -4.2188        | 1.0078          | -5.2188          |
| 0.0239        | 2.5118 | 2400 | -10.1875      | -8.875          | -736.0       | -800.0         | 0.6858          | 0.7383             | -4.1562        | 0.9766          | -5.125           |
| 0.0188        | 2.6164 | 2500 | -10.125       | -8.8125         | -744.0       | -816.0         | 0.6902          | 0.7324             | -4.2812        | 0.9883          | -5.25            |
| 0.0145        | 2.7211 | 2600 | -10.125       | -8.8125         | -748.0       | -816.0         | 0.6874          | 0.7383             | -4.2812        | 0.9844          | -5.2812          |
| 0.0229        | 2.8257 | 2700 | -10.0625      | -8.75           | -752.0       | -824.0         | 0.6883          | 0.7344             | -4.3438        | 0.9922          | -5.3438          |
| 0.0298        | 2.9304 | 2800 | -10.0625      | -8.75           | -752.0       | -824.0         | 0.6883          | 0.7344             | -4.3438        | 0.9922          | -5.3438          |


### Framework versions

- Transformers 4.44.2
- Pytorch 2.3.0
- Datasets 3.0.0
- Tokenizers 0.19.1