Modelpublic

hiwden00/whisper-multiclass-lang-en-base

sourceHugging Faceapache-2.0updated 2y agoView on Hugging Face

0likes10downloads

Model Card

whisper-multiclass-lang-en-base

This model is a fine-tuned version of openai/whisper-base on an unknown dataset. It achieves the following results on the evaluation set:

—Loss: 0.1493
—Wer: 6.2678
—Cer: 4.3198

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: 500
—training_steps: 5000
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.0286	4.5872	500	0.1880	10.4938	6.6816
0.0011	9.1743	1000	0.1529	7.6923	5.1185
0.0003	13.7615	1500	0.1525	7.1700	4.8523
0.0002	18.3486	2000	0.1510	7.0275	4.8695
0.0001	22.9358	2500	0.1505	6.7426	4.5946
0.0001	27.5229	3000	0.1499	6.6952	4.5861
0.0001	32.1101	3500	0.1496	6.4103	4.4143
0.0001	36.6972	4000	0.1495	6.5527	4.4830
0.0001	41.2844	4500	0.1493	6.3153	4.3456
0.0001	45.8716	5000	0.1493	6.2678	4.3198

Framework versions

—Transformers 4.45.1
—Pytorch 2.4.1+cu121
—Datasets 3.0.1
—Tokenizers 0.20.0