File size: 5,275 Bytes
94cecc8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: llama3
base_model: RyanYr/openchat-3.6-8b-20240522_iter1
tags:
- trl
- dpo
- generated_from_trainer
model-index:
- name: openchat-3.6-8b-20240522_iter2
  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. -->

# openchat-3.6-8b-20240522_iter2

This model is a fine-tuned version of [RyanYr/openchat-3.6-8b-20240522_iter1](https://huggingface.co/RyanYr/openchat-3.6-8b-20240522_iter1) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.5377
- Rewards/chosen: -0.2459
- Rewards/rejected: -0.7375
- Rewards/accuracies: 0.7200
- Rewards/margins: 0.4916
- Logps/rejected: -139.4380
- Logps/chosen: -132.3574
- Logits/rejected: -1.3194
- Logits/chosen: -1.3369

## 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: 2e-07
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 16
- total_train_batch_size: 64
- total_eval_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 100
- num_epochs: 2

### Training results

| Training Loss | Epoch  | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
| 0.7344        | 0.1140 | 100  | 0.7121          | 0.3107         | 0.3572           | 0.4000             | -0.0465         | -128.4905      | -126.7911    | -1.4523         | -1.4659       |
| 0.6553        | 0.2281 | 200  | 0.6966          | 0.2237         | 0.2513           | 0.5200             | -0.0276         | -129.5494      | -127.6609    | -1.4428         | -1.4571       |
| 0.669         | 0.3421 | 300  | 0.6798          | 0.1031         | 0.0581           | 0.5200             | 0.0450          | -131.4815      | -128.8666    | -1.4122         | -1.4281       |
| 0.6402        | 0.4561 | 400  | 0.6595          | 0.0694         | -0.0114          | 0.6400             | 0.0808          | -132.1772      | -129.2041    | -1.4254         | -1.4406       |
| 0.6716        | 0.5702 | 500  | 0.6351          | 0.1022         | -0.0221          | 0.6400             | 0.1243          | -132.2838      | -128.8764    | -1.4550         | -1.4689       |
| 0.655         | 0.6842 | 600  | 0.6278          | 0.1039         | -0.0286          | 0.6000             | 0.1325          | -132.3487      | -128.8587    | -1.4625         | -1.4766       |
| 0.5943        | 0.7982 | 700  | 0.6084          | 0.0643         | -0.1073          | 0.6400             | 0.1716          | -133.1360      | -129.2548    | -1.4485         | -1.4622       |
| 0.6048        | 0.9123 | 800  | 0.6002          | 0.0902         | -0.1175          | 0.6800             | 0.2077          | -133.2379      | -128.9962    | -1.4607         | -1.4735       |
| 0.4934        | 1.0263 | 900  | 0.5798          | 0.0298         | -0.2745          | 0.7200             | 0.3043          | -134.8078      | -129.5996    | -1.4349         | -1.4491       |
| 0.4284        | 1.1403 | 1000 | 0.5724          | -0.1252        | -0.4897          | 0.6800             | 0.3645          | -136.9601      | -131.1501    | -1.3824         | -1.3981       |
| 0.4132        | 1.2544 | 1100 | 0.5563          | -0.1930        | -0.5928          | 0.7600             | 0.3998          | -137.9906      | -131.8278    | -1.3545         | -1.3715       |
| 0.3957        | 1.3684 | 1200 | 0.5543          | -0.2162        | -0.6427          | 0.7600             | 0.4264          | -138.4894      | -132.0604    | -1.3412         | -1.3583       |
| 0.4893        | 1.4824 | 1300 | 0.5476          | -0.2078        | -0.6782          | 0.7200             | 0.4704          | -138.8445      | -131.9757    | -1.3340         | -1.3521       |
| 0.4361        | 1.5965 | 1400 | 0.5413          | -0.2007        | -0.6908          | 0.7200             | 0.4901          | -138.9703      | -131.9046    | -1.3316         | -1.3490       |
| 0.4406        | 1.7105 | 1500 | 0.5477          | -0.2466        | -0.6913          | 0.7200             | 0.4448          | -138.9762      | -132.3638    | -1.3242         | -1.3421       |
| 0.3988        | 1.8245 | 1600 | 0.5449          | -0.2388        | -0.7225          | 0.7200             | 0.4838          | -139.2881      | -132.2855    | -1.3254         | -1.3431       |
| 0.4044        | 1.9386 | 1700 | 0.5377          | -0.2459        | -0.7375          | 0.7200             | 0.4916          | -139.4380      | -132.3574    | -1.3194         | -1.3369       |


### Framework versions

- Transformers 4.43.4
- Pytorch 2.1.2+cu121
- Datasets 2.14.6
- Tokenizers 0.19.1