Commit
·
9228af8
1
Parent(s):
3d54b7e
End of training
Browse files
README.md
ADDED
|
@@ -0,0 +1,87 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
base_model: Harveenchadha/vakyansh-wav2vec2-hindi-him-4200
|
| 4 |
+
tags:
|
| 5 |
+
- generated_from_trainer
|
| 6 |
+
datasets:
|
| 7 |
+
- fleurs
|
| 8 |
+
metrics:
|
| 9 |
+
- wer
|
| 10 |
+
model-index:
|
| 11 |
+
- name: hindi_wav2vec2_optimized
|
| 12 |
+
results:
|
| 13 |
+
- task:
|
| 14 |
+
name: Automatic Speech Recognition
|
| 15 |
+
type: automatic-speech-recognition
|
| 16 |
+
dataset:
|
| 17 |
+
name: fleurs
|
| 18 |
+
type: fleurs
|
| 19 |
+
config: hi_in
|
| 20 |
+
split: test
|
| 21 |
+
args: hi_in
|
| 22 |
+
metrics:
|
| 23 |
+
- name: Wer
|
| 24 |
+
type: wer
|
| 25 |
+
value: 0.4386410231377526
|
| 26 |
+
---
|
| 27 |
+
|
| 28 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 29 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 30 |
+
|
| 31 |
+
# hindi_wav2vec2_optimized
|
| 32 |
+
|
| 33 |
+
This model is a fine-tuned version of [Harveenchadha/vakyansh-wav2vec2-hindi-him-4200](https://huggingface.co/Harveenchadha/vakyansh-wav2vec2-hindi-him-4200) on the fleurs dataset.
|
| 34 |
+
It achieves the following results on the evaluation set:
|
| 35 |
+
- Loss: 2.2950
|
| 36 |
+
- Wer: 0.4386
|
| 37 |
+
|
| 38 |
+
## Model description
|
| 39 |
+
|
| 40 |
+
More information needed
|
| 41 |
+
|
| 42 |
+
## Intended uses & limitations
|
| 43 |
+
|
| 44 |
+
More information needed
|
| 45 |
+
|
| 46 |
+
## Training and evaluation data
|
| 47 |
+
|
| 48 |
+
More information needed
|
| 49 |
+
|
| 50 |
+
## Training procedure
|
| 51 |
+
|
| 52 |
+
### Training hyperparameters
|
| 53 |
+
|
| 54 |
+
The following hyperparameters were used during training:
|
| 55 |
+
- learning_rate: 0.0003
|
| 56 |
+
- train_batch_size: 16
|
| 57 |
+
- eval_batch_size: 8
|
| 58 |
+
- seed: 42
|
| 59 |
+
- gradient_accumulation_steps: 2
|
| 60 |
+
- total_train_batch_size: 32
|
| 61 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 62 |
+
- lr_scheduler_type: linear
|
| 63 |
+
- lr_scheduler_warmup_steps: 10
|
| 64 |
+
- num_epochs: 125
|
| 65 |
+
|
| 66 |
+
### Training results
|
| 67 |
+
|
| 68 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
| 69 |
+
|:-------------:|:-----:|:----:|:---------------:|:------:|
|
| 70 |
+
| 3.8474 | 10.0 | 25 | 2.0026 | 0.5689 |
|
| 71 |
+
| 0.5389 | 20.0 | 50 | 2.0278 | 0.4474 |
|
| 72 |
+
| 0.4199 | 30.0 | 75 | 2.1555 | 0.4512 |
|
| 73 |
+
| 0.1534 | 40.0 | 100 | 2.2412 | 0.4766 |
|
| 74 |
+
| 0.096 | 50.0 | 125 | 2.2013 | 0.4593 |
|
| 75 |
+
| 0.0574 | 60.0 | 150 | 2.1777 | 0.4657 |
|
| 76 |
+
| 0.0361 | 70.0 | 175 | 2.4166 | 0.4391 |
|
| 77 |
+
| 0.0278 | 80.0 | 200 | 2.2826 | 0.4571 |
|
| 78 |
+
| 0.025 | 90.0 | 225 | 2.3623 | 0.4433 |
|
| 79 |
+
| 0.0157 | 100.0 | 250 | 2.2950 | 0.4386 |
|
| 80 |
+
|
| 81 |
+
|
| 82 |
+
### Framework versions
|
| 83 |
+
|
| 84 |
+
- Transformers 4.34.0
|
| 85 |
+
- Pytorch 2.0.1+cu118
|
| 86 |
+
- Datasets 2.14.5
|
| 87 |
+
- Tokenizers 0.14.1
|