parambharat/whisper-tiny-ta

Whisper Tiny Ta - Bharat Ramanathan

This model is a fine-tuned version of openai/whisper-tiny on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3096
• Wer: 30.1027

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: 1e-05
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5622	0.2	1000	0.4460	41.4141
0.4151	0.4	2000	0.3657	35.1390
0.3727	0.6	3000	0.3417	33.1723
0.3519	0.8	4000	0.3252	31.9497
0.3354	1.0	5000	0.3192	31.3997
0.3492	0.1	6000	0.3283	31.6966
0.3229	0.2	7000	0.3211	31.1339
0.3193	0.3	8000	0.3138	30.5161
0.314	0.4	9000	0.3112	30.1832
0.3087	0.5	10000	0.3096	30.1027

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2