Modelpublic

Badri0510/whisper-tanglish-ultra-8000

sourceHugging Faceupdated 5mo agoView on Hugging Face

1likes30downloads

Model Card

whisper-tanglish-ultra-8000

This model is a fine-tuned version of Badri0510/whisper-tanglish-voice-generalized-5000 on an unknown dataset. It achieves the following results on the evaluation set:

—Loss: 0.4485
—Wer: 0.5705

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: 0.0001
—trainbatchsize: 8
—evalbatchsize: 8
—seed: 42
—gradientaccumulationsteps: 2
—totaltrainbatch_size: 16
—optimizer: Use OptimizerNames.ADAMWTORCHFUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
—lrschedulertype: constant
—training_steps: 8000
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2356	0.0625	500	0.4658	0.5775
0.2673	1.0137	1000	0.4256	0.5774
0.2367	1.0762	1500	0.4124	0.5712
0.1790	2.0264	2000	0.4188	0.5662
0.1995	2.0889	2500	0.4006	0.5732
0.1951	3.0395	3000	0.4093	0.5681
0.2177	3.102	3500	0.4032	0.5706
0.1388	4.0533	4000	0.4216	0.5706
0.1281	5.0046	4500	0.4355	0.5786
0.1366	5.0671	5000	0.4171	0.5695
0.1010	6.0186	5500	0.4326	0.5703
0.1213	6.0811	6000	0.4374	0.5725
0.0991	7.0315	6500	0.4540	0.5745
0.1052	7.094	7000	0.4337	0.5697
0.1039	8.0451	7500	0.4511	0.5721
0.1139	8.1076	8000	0.4485	0.5705

Framework versions

—Transformers 5.0.0
—Pytorch 2.10.0+cu128
—Datasets 4.0.0
—Tokenizers 0.22.2