File size: 2,850 Bytes
30316ca
7c77e8f
 
 
 
 
 
 
 
 
30316ca
 
7c77e8f
 
30316ca
7c77e8f
30316ca
7c77e8f
 
 
 
 
 
 
 
 
 
 
 
 
 
30316ca
7c77e8f
30316ca
7c77e8f
30316ca
7c77e8f
30316ca
7c77e8f
30316ca
7c77e8f
30316ca
7c77e8f
30316ca
7c77e8f
30316ca
7c77e8f
30316ca
7c77e8f
 
 
 
 
 
 
 
 
 
 
30316ca
7c77e8f
30316ca
7c77e8f
 
 
 
 
30316ca
 
7c77e8f
30316ca
7c77e8f
 
 
 
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
---
license: apache-2.0
base_model: HuggingFaceTB/SmolLM-135M-Instruct
tags:
- trl
- orpo
- generated_from_trainer
model-index:
- name: ft-orpo-smollm-135M-instruct-on-hf-ultrafeedback
  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. -->

# ft-orpo-smollm-135M-instruct-on-hf-ultrafeedback

This model is a fine-tuned version of [HuggingFaceTB/SmolLM-135M-Instruct](https://huggingface.co/HuggingFaceTB/SmolLM-135M-Instruct) on the HuggingFaceH4/ultrafeedback_binarized dataset.
It achieves the following results on the evaluation set:
- Loss: 1.1646
- Rewards/chosen: -0.1296
- Rewards/rejected: -0.1298
- Rewards/accuracies: 0.4000
- Rewards/margins: 0.0002
- Logps/rejected: -1.2981
- Logps/chosen: -1.2964
- Logits/rejected: 31.6875
- Logits/chosen: 31.3425
- Nll Loss: 1.0873
- Log Odds Ratio: -0.7727
- Log Odds Chosen: -0.0238

## 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: 0.0003
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 1

### 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 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|:--------:|:--------------:|:---------------:|
| 1.4274        | 0.27  | 100  | 1.2066          | -0.1351        | -0.1347          | 0.4100             | -0.0004         | -1.3467        | -1.3508      | 28.6347         | 28.3442       | 1.1292   | -0.7736        | -0.0347         |
| 1.1351        | 0.53  | 200  | 1.1796          | -0.1316        | -0.1316          | 0.4100             | 0.0000          | -1.3162        | -1.3158      | 31.1292         | 30.7764       | 1.1024   | -0.7723        | -0.0251         |
| 1.135         | 0.8   | 300  | 1.1646          | -0.1296        | -0.1298          | 0.4000             | 0.0002          | -1.2981        | -1.2964      | 31.6875         | 31.3425       | 1.0873   | -0.7727        | -0.0238         |


### Framework versions

- Transformers 4.39.3
- Pytorch 2.1.2
- Datasets 2.18.0
- Tokenizers 0.15.2