|
--- |
|
tags: |
|
- generated_from_trainer |
|
datasets: |
|
- common_voice |
|
model-index: |
|
- name: wav2vec2-60-urdu |
|
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. --> |
|
|
|
# wav2vec2-60-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: 6.4496 |
|
- Wer: 0.5913 |
|
- Cer: 0.3310 |
|
|
|
## 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: 0.0003 |
|
- 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 | |
|
|:-------------:|:-----:|:----:|:---------------:|:------:|:------:| |
|
| 12.6045 | 8.33 | 100 | 8.4997 | 0.6978 | 0.3923 | |
|
| 1.3367 | 16.67 | 200 | 5.0015 | 0.6515 | 0.3556 | |
|
| 0.5344 | 25.0 | 300 | 9.3687 | 0.6393 | 0.3625 | |
|
| 0.2922 | 33.33 | 400 | 9.2381 | 0.6236 | 0.3432 | |
|
| 0.1867 | 41.67 | 500 | 6.2150 | 0.6035 | 0.3448 | |
|
| 0.1166 | 50.0 | 600 | 6.4496 | 0.5913 | 0.3310 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.15.0 |
|
- Pytorch 1.10.0+cu111 |
|
- Datasets 1.17.0 |
|
- Tokenizers 0.10.3 |
|
|
