File size: 1,682 Bytes
59220ca
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9e4ec58
 
59220ca
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9e4ec58
 
 
 
59220ca
 
 
 
 
 
 
 
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
---
license: apache-2.0
base_model: jonatasgrosman/wav2vec2-large-xlsr-53-arabic
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: tamasheq-5
  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. -->

# tamasheq-5

This model is a fine-tuned version of [jonatasgrosman/wav2vec2-large-xlsr-53-arabic](https://huggingface.co/jonatasgrosman/wav2vec2-large-xlsr-53-arabic) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.2985
- Wer: 0.4870

## 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: 3e-05
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 30

### Training results

| Training Loss | Epoch | Step | Validation Loss | Wer    |
|:-------------:|:-----:|:----:|:---------------:|:------:|
| 10.0913       | 6.06  | 400  | 3.0405          | 1.0    |
| 1.2281        | 12.12 | 800  | 0.3682          | 0.5056 |
| 0.3094        | 18.18 | 1200 | 0.3076          | 0.4878 |
| 0.2389        | 24.24 | 1600 | 0.2985          | 0.4870 |


### Framework versions

- Transformers 4.31.0
- Pytorch 2.0.1+cu118
- Datasets 2.14.4
- Tokenizers 0.13.3