Modelpublic

golesheed/whisper-v2-North-Brabantic_and_river_area_Guelders

sourceHugging Faceapache-2.0updated 2y agoView on Hugging Face

0likes3downloads

Model Card

Whisper Large V2

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: 0.5449
—Wer: 25.0953

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: 3e-05
—trainbatchsize: 12
—evalbatchsize: 8
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 20
—num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.841	0.2239	15	0.5933	48.4347
0.5965	0.4478	30	0.5012	27.9249
0.5018	0.6716	45	0.4670	25.0251
0.4578	0.8955	60	0.4569	27.6390
0.3824	1.1194	75	0.4603	27.2978
0.2738	1.3433	90	0.4537	25.0301
0.2375	1.5672	105	0.4516	24.4632
0.2573	1.7910	120	0.4381	25.3512
0.241	2.0149	135	0.4379	25.4766
0.1265	2.2388	150	0.4624	23.7809
0.1391	2.4627	165	0.4588	26.6406
0.1242	2.6866	180	0.4572	24.7642
0.1227	2.9104	195	0.4561	27.5738
0.0774	3.1343	210	0.4790	24.2474
0.0543	3.3582	225	0.4931	31.8483
0.0506	3.5821	240	0.5087	25.3010
0.056	3.8060	255	0.4933	27.6942
0.0527	4.0299	270	0.5009	26.2543
0.0233	4.2537	285	0.5447	27.8999
0.0193	4.4776	300	0.5458	27.0570
0.0167	4.7015	315	0.5421	24.5384
0.0183	4.9254	330	0.5449	25.0953

Framework versions

—Transformers 4.45.0.dev0
—Pytorch 2.1.0+cu121
—Datasets 2.20.0
—Tokenizers 0.19.1