Modelpublic

bagasshw/whisper-large-v2-javanese-openslr-v1

sourceHugging Faceapache-2.0updated 1y agoView on Hugging Face

0likes54downloads

Model Card

Whisper-Large-V2-Java-v1

This model is a fine-tuned version of openai/whisper-large-v2 on the None dataset. It achieves the following results on the evaluation set:

—Loss: 0.1591
—Wer: 0.1198

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: 2e-05
—trainbatchsize: 2
—evalbatchsize: 2
—seed: 42
—gradientaccumulationsteps: 2
—totaltrainbatch_size: 4
—optimizer: Use OptimizerNames.ADAMWTORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizerargs=No additional optimizer arguments
—lrschedulertype: linear
—lrschedulerwarmup_ratio: 0.1
—training_steps: 20000
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4753	0.0541	1000	0.4285	0.3084
0.413	0.1081	2000	0.3664	0.2545
0.4282	0.1622	3000	0.3684	0.2546
0.4007	0.2163	4000	0.3713	0.2529
0.4538	0.2703	5000	0.3791	0.2584
0.4111	0.3244	6000	0.3602	0.2457
0.3969	0.3785	7000	0.3437	0.2315
0.3393	0.4325	8000	0.3217	0.2244
0.3356	0.4866	9000	0.2984	0.2049
0.3161	0.5407	10000	0.2903	0.2001
0.3198	0.5947	11000	0.2882	0.2004
0.3023	0.6488	12000	0.2731	0.1889
0.2368	0.7029	13000	0.2201	0.1572
0.2348	0.7569	14000	0.2022	0.1488
0.2062	0.8110	15000	0.1923	0.1403
0.1859	0.8651	16000	0.1798	0.1331
0.1808	0.9191	17000	0.1700	0.1283
0.1689	0.9732	18000	0.1635	0.1253
0.0984	1.0272	19000	0.1606	0.1211
0.0853	1.0813	20000	0.1591	0.1198

Framework versions

—Transformers 4.50.0.dev0
—Pytorch 2.6.0+cu126
—Datasets 2.16.0
—Tokenizers 0.21.1