bnriiitb/whisper-small-te

Whisper Small Telugu - Naga Budigam

This model is a fine-tuned version of openai/whisper-small on the Chai_Bisket_Stories_16-08-2021_14-17 dataset. It achieves the following results on the evaluation set:

• Loss: 0.7063
• Wer: 77.4871

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: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2933	2.62	500	0.3849	86.6429
0.0692	5.24	1000	0.3943	82.7190
0.0251	7.85	1500	0.4720	82.4415
0.0098	10.47	2000	0.5359	81.6092
0.0061	13.09	2500	0.5868	75.9413
0.0025	15.71	3000	0.6235	76.6944
0.0009	18.32	3500	0.6634	78.3987
0.0005	20.94	4000	0.6776	77.1700
0.0002	23.56	4500	0.6995	78.2798
0.0001	26.18	5000	0.7063	77.4871

Framework versions

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