File size: 3,821 Bytes
873a17c
a04c2fe
 
 
873a17c
a04c2fe
 
873a17c
 
a04c2fe
 
 
873a17c
 
a04c2fe
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
873a17c
 
 
 
 
 
 
7c41b61
873a17c
7c41b61
 
 
 
 
 
a04c2fe
873a17c
7c41b61
873a17c
a04c2fe
873a17c
 
 
 
 
 
 
 
 
 
 
 
 
 
d49441f
 
873a17c
 
 
 
d49441f
 
7c41b61
 
 
 
 
 
873a17c
 
 
 
d49441f
873a17c
 
 
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
102
103
104
105
106
107
108
109
110
---
language: 
- ur
license: apache-2.0
tags:
- automatic-speech-recognition
- robust-speech-event
datasets:
- common_voice
metrics:
- wer
- cer
model-index:
- name: wav2vec2-large-xlsr-53-urdu
  results:
  - task: 
      type: automatic-speech-recognition  # Required. Example: automatic-speech-recognition
      name: Urdu Speech Recognition  # Optional. Example: Speech Recognition
    dataset:
      type: common_voice  # Required. Example: common_voice. Use dataset id from https://hf.co/datasets
      name: Urdu  # Required. Example: Common Voice zh-CN
      args: ur         # Optional. Example: zh-CN
    metrics:
      - type: wer    # Required. Example: wer
        value: 57.7  # Required. Example: 20.90
        name: Test WER    # Optional. Example: Test WER
        args: 
        - learning_rate: 0.0003
        - 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: 200
        - num_epochs: 50
        - mixed_precision_training: Native AMP         # Optional. Example for BLEU: max_order
      - type: cer    # Required. Example: wer
        value: 33.8  # Required. Example: 20.90
        name: Test CER    # Optional. Example: Test WER
        args: 
        - learning_rate: 0.0003
        - 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: 200
        - num_epochs: 50
        - mixed_precision_training: Native AMP         # Optional. Example for BLEU: max_order
---

<!-- 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. -->

# wav2vec2-large-xlsr-53-urdu

This model is a fine-tuned version of [Harveenchadha/vakyansh-wav2vec2-urdu-urm-60](https://huggingface.co/Harveenchadha/vakyansh-wav2vec2-urdu-urm-60) on the common_voice dataset.
It achieves the following results on the evaluation set:
- Loss: 11.4593
- Wer: 0.5772
- Cer: 0.3384

## Model description

The training and valid dataset is 0.58 hours. It was hard to train any model on lower number of so I decided to take Urdu checkpoint and finetune the XLSR model.

## Training and evaluation data

Trained on Harveenchadha/vakyansh-wav2vec2-urdu-urm-60 due to lesser number of samples. Persian and Urdu are quite similar. 

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 0.0003
- 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: 200
- num_epochs: 50
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch | Step | Validation Loss | Wer    | Cer    |
|:-------------:|:-----:|:----:|:---------------:|:------:|:------:|
| 13.2136       | 8.33  | 100  | 9.5424          | 0.7672 | 0.4381 |
| 2.6996        | 16.67 | 200  | 8.4317          | 0.6661 | 0.3620 |
| 1.371         | 25.0  | 300  | 9.5518          | 0.6443 | 0.3701 |
| 0.639         | 33.33 | 400  | 9.4132          | 0.6129 | 0.3609 |
| 0.4452        | 41.67 | 500  | 10.8330         | 0.5920 | 0.3473 |
| 0.3233        | 50.0  | 600  | 11.4593         | 0.5772 | 0.3384 |


### Framework versions

- Transformers 4.15.0
- Pytorch 1.10.0+cu111
- Datasets 1.17.0
- Tokenizers 0.10.3