Hugging Face's logo

shivam
/
wav2vec2-xls-r-300m-hindi

Automatic Speech Recognition
Transformers
PyTorch
Hindi
wav2vec2
mozilla-foundation/common_voice_7_0
Generated from Trainer
Model card Files
xet
 Community
1
shivam commited on
Commit
3384b3c
·
1 Parent(s): 7554811

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +95 -0
README.md ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - hi
4
+ license: apache-2.0
5
+ tags:
6
+ - automatic-speech-recognition
7
+ - mozilla-foundation/common_voice_7_0
8
+ - generated_from_trainer
9
+ datasets:
10
+ - common_voice
11
+ model-index:
12
+ - name: ''
13
+ results: []
14
+ ---
15
+
16
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
17
+ should probably proofread and complete it, then remove this comment. -->
18
+
19
+ #
20
+
21
+ This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the MOZILLA-FOUNDATION/COMMON_VOICE_7_0 - HI dataset.
22
+ It achieves the following results on the evaluation set:
23
+ - Loss: 1.4031
24
+ - Wer: 0.6827
25
+
26
+ ## Model description
27
+
28
+ More information needed
29
+
30
+ ## Intended uses & limitations
31
+
32
+ More information needed
33
+
34
+ ## Training and evaluation data
35
+
36
+ More information needed
37
+
38
+ ## Training procedure
39
+
40
+ ### Training hyperparameters
41
+
42
+ The following hyperparameters were used during training:
43
+ - learning_rate: 7.5e-05
44
+ - train_batch_size: 8
45
+ - eval_batch_size: 8
46
+ - seed: 42
47
+ - gradient_accumulation_steps: 4
48
+ - total_train_batch_size: 32
49
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
50
+ - lr_scheduler_type: linear
51
+ - lr_scheduler_warmup_steps: 2000
52
+ - num_epochs: 100.0
53
+ - mixed_precision_training: Native AMP
54
+
55
+ ### Training results
56
+
57
+ | Training Loss | Epoch | Step | Validation Loss | Wer |
58
+ |:-------------:|:-----:|:-----:|:---------------:|:------:|
59
+ | 5.3156 | 3.4 | 500 | 4.5583 | 1.0 |
60
+ | 3.3329 | 6.8 | 1000 | 3.4274 | 1.0001 |
61
+ | 2.1275 | 10.2 | 1500 | 1.7221 | 0.8763 |
62
+ | 1.5737 | 13.6 | 2000 | 1.4188 | 0.8143 |
63
+ | 1.3835 | 17.01 | 2500 | 1.2251 | 0.7447 |
64
+ | 1.3247 | 20.41 | 3000 | 1.2827 | 0.7394 |
65
+ | 1.231 | 23.81 | 3500 | 1.2216 | 0.7074 |
66
+ | 1.1819 | 27.21 | 4000 | 1.2210 | 0.6863 |
67
+ | 1.1546 | 30.61 | 4500 | 1.3233 | 0.7308 |
68
+ | 1.0902 | 34.01 | 5000 | 1.3251 | 0.7010 |
69
+ | 1.0749 | 37.41 | 5500 | 1.3274 | 0.7235 |
70
+ | 1.0412 | 40.81 | 6000 | 1.2942 | 0.6856 |
71
+ | 1.0064 | 44.22 | 6500 | 1.2581 | 0.6732 |
72
+ | 1.0006 | 47.62 | 7000 | 1.2767 | 0.6885 |
73
+ | 0.9518 | 51.02 | 7500 | 1.2966 | 0.6925 |
74
+ | 0.9514 | 54.42 | 8000 | 1.2981 | 0.7067 |
75
+ | 0.9241 | 57.82 | 8500 | 1.3835 | 0.7124 |
76
+ | 0.9059 | 61.22 | 9000 | 1.3318 | 0.7083 |
77
+ | 0.8906 | 64.62 | 9500 | 1.3640 | 0.6962 |
78
+ | 0.8468 | 68.03 | 10000 | 1.4727 | 0.6982 |
79
+ | 0.8631 | 71.43 | 10500 | 1.3401 | 0.6809 |
80
+ | 0.8154 | 74.83 | 11000 | 1.4124 | 0.6955 |
81
+ | 0.7953 | 78.23 | 11500 | 1.4245 | 0.6950 |
82
+ | 0.818 | 81.63 | 12000 | 1.3944 | 0.6995 |
83
+ | 0.7772 | 85.03 | 12500 | 1.3735 | 0.6785 |
84
+ | 0.7857 | 88.43 | 13000 | 1.3696 | 0.6808 |
85
+ | 0.7705 | 91.84 | 13500 | 1.4101 | 0.6870 |
86
+ | 0.7537 | 95.24 | 14000 | 1.4178 | 0.6832 |
87
+ | 0.7734 | 98.64 | 14500 | 1.4027 | 0.6831 |
88
+
89
+
90
+ ### Framework versions
91
+
92
+ - Transformers 4.16.0.dev0
93
+ - Pytorch 1.10.1+cu113
94
+ - Datasets 1.18.1.dev0
95
+ - Tokenizers 0.11.0