metadata
base_model: Vignesh-M/Indic-whisper
tags:
- generated_from_trainer
metrics:
- accuracy
model-index:
- name: whisper-indic-audio-abuse-feature
results: []
whisper-indic-audio-abuse-feature
This model is a fine-tuned version of Vignesh-M/Indic-whisper on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.5056
- Accuracy: 0.8868
- Macro Precision: 0.8642
- Macro Recall: 0.8509
- Macro F1-score: 0.8572
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: 5e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.01
- num_epochs: 5
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Macro Precision | Macro Recall | Macro F1-score |
|---|---|---|---|---|---|---|---|
| 0.4635 | 0.4367 | 50 | 0.4010 | 0.8020 | 0.8176 | 0.8096 | 0.8014 |
| 0.3403 | 0.8734 | 100 | 0.3162 | 0.8684 | 0.8685 | 0.8668 | 0.8675 |
| 0.2689 | 1.3100 | 150 | 0.3025 | 0.8807 | 0.8838 | 0.8774 | 0.8793 |
| 0.2339 | 1.7467 | 200 | 0.3019 | 0.8782 | 0.8776 | 0.8777 | 0.8776 |
| 0.1723 | 2.1834 | 250 | 0.3715 | 0.8868 | 0.8870 | 0.8854 | 0.8861 |
| 0.1027 | 2.6201 | 300 | 0.3472 | 0.8930 | 0.8937 | 0.8912 | 0.8921 |
| 0.123 | 3.0568 | 350 | 0.3690 | 0.8795 | 0.8855 | 0.8751 | 0.8776 |
| 0.0497 | 3.4934 | 400 | 0.4423 | 0.8918 | 0.8916 | 0.8907 | 0.8911 |
| 0.0534 | 3.9301 | 450 | 0.3937 | 0.9041 | 0.9048 | 0.9024 | 0.9033 |
| 0.0235 | 4.3668 | 500 | 0.4753 | 0.8979 | 0.8993 | 0.8958 | 0.8970 |
| 0.0196 | 4.8035 | 550 | 0.5204 | 0.8967 | 0.8982 | 0.8944 | 0.8957 |
Framework versions
- Transformers 4.41.2
- Pytorch 2.1.2
- Datasets 2.19.2
- Tokenizers 0.19.1