Modelpublic

ThangaTharun/finetunedtiny

sourceHugging Faceapache-2.0updated 3y agoView on Hugging Face

0likes9downloads

Model Card

ThangaTharun/Bharishkafinetuned

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

—Loss: 1.6092
—Wer: 0.0

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
—trainbatchsize: 16
—evalbatchsize: 8
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 1
—training_steps: 60
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
6.0126	1.67	5	5.7191	27.2727
4.5556	3.33	10	4.0076	18.1818
3.385	5.0	15	3.0201	22.7273
2.7403	6.67	20	2.2411	18.1818
1.9444	8.33	25	2.0404	4.5455
1.9864	10.0	30	1.9273	4.5455
2.3372	11.67	35	1.8292	0.0
1.5928	13.33	40	1.7561	0.0
1.8158	15.0	45	1.6984	0.0
1.7321	16.67	50	1.6561	0.0
1.7266	18.33	55	1.6268	0.0
1.6997	20.0	60	1.6092	0.0

Framework versions

—Transformers 4.36.2
—Pytorch 2.1.0+cu121
—Datasets 2.16.1
—Tokenizers 0.15.0