File size: 3,581 Bytes
ea7acb7
 
 
33c380c
ea7acb7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
33c380c
 
 
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
---
tags:
- generated_from_trainer
base_model: Lakoc/DeCRED_small_cv_2
datasets:
- common_voice_13_0
metrics:
- wer
model-index:
- name: DeCRED_linear_mixing_tuning
  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. -->

# DeCRED_linear_mixing_tuning

This model is a fine-tuned version of [Lakoc/DeCRED_small_cv_2](https://huggingface.co/Lakoc/DeCRED_small_cv_2) on the common_voice_13_0 dataset.
It achieves the following results on the evaluation set:
- Loss: 1.0590
- Cer: 0.0632
- Wer: 0.1471
- Mer: 0.1444
- Wil: 0.2408
- Wip: 0.7592
- Hits: 23158
- Substitutions: 2931
- Deletions: 484
- Insertions: 494

## 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: 0.05
- train_batch_size: 256
- eval_batch_size: 64
- seed: 42
- distributed_type: multi-GPU
- gradient_accumulation_steps: 4
- total_train_batch_size: 1024
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 50.0

### Training results

| Training Loss | Epoch | Step | Validation Loss | Cer    | Wer    | Mer    | Wil    | Wip    | Hits  | Substitutions | Deletions | Insertions |
|:-------------:|:-----:|:----:|:---------------:|:------:|:------:|:------:|:------:|:------:|:-----:|:-------------:|:---------:|:----------:|
| 1.185         | 2.67  | 20   | 1.1215          | 0.0646 | 0.1520 | 0.1492 | 0.2479 | 0.7521 | 23030 | 3013          | 530       | 496        |
| 1.0639        | 5.33  | 40   | 1.0717          | 0.0634 | 0.1480 | 0.1453 | 0.2420 | 0.7580 | 23124 | 2939          | 510       | 483        |
| 1.0957        | 8.0   | 60   | 1.0612          | 0.0628 | 0.1464 | 0.1438 | 0.2402 | 0.7598 | 23162 | 2929          | 482       | 480        |
| 1.0936        | 10.67 | 80   | 1.0595          | 0.0631 | 0.1469 | 0.1442 | 0.2407 | 0.7593 | 23158 | 2934          | 481       | 488        |
| 1.0804        | 13.33 | 100  | 1.0591          | 0.0630 | 0.1468 | 0.1441 | 0.2405 | 0.7595 | 23164 | 2929          | 480       | 492        |
| 1.1044        | 16.0  | 120  | 1.0591          | 0.0631 | 0.1469 | 0.1442 | 0.2405 | 0.7595 | 23162 | 2929          | 482       | 492        |
| 1.0836        | 18.67 | 140  | 1.0589          | 0.0631 | 0.1468 | 0.1442 | 0.2405 | 0.7595 | 23163 | 2929          | 481       | 492        |
| 1.0924        | 21.33 | 160  | 1.0590          | 0.0632 | 0.1471 | 0.1444 | 0.2408 | 0.7592 | 23158 | 2931          | 484       | 494        |
| 1.1048        | 24.0  | 180  | 1.0590          | 0.0632 | 0.1471 | 0.1444 | 0.2408 | 0.7592 | 23158 | 2931          | 484       | 494        |
| 1.0858        | 26.67 | 200  | 1.0589          | 0.0631 | 0.1470 | 0.1443 | 0.2407 | 0.7593 | 23160 | 2929          | 484       | 494        |
| 1.0953        | 29.33 | 220  | 1.0589          | 0.0631 | 0.1468 | 0.1442 | 0.2405 | 0.7595 | 23163 | 2929          | 481       | 492        |
| 1.1308        | 32.0  | 240  | 1.0590          | 0.0632 | 0.1471 | 0.1444 | 0.2408 | 0.7592 | 23158 | 2931          | 484       | 494        |


### Framework versions

- Transformers 4.40.0.dev0
- Pytorch 2.2.0+rocm5.6
- Datasets 2.18.0
- Tokenizers 0.15.2

### Wandb run
https://wandb.ai/butspeechfit/decred_commonvoice_en/runs/DeCRED_linear_mixing_tuning