Modelpublic

Gummybear05/whisper-small-E50_freq_speed2

sourceHugging Faceapache-2.0updated 2y agoView on Hugging Face

0likes6downloads

Model Card

whisper-small-E50Yfreqspeed

This model is a fine-tuned version of openai/whisper-small on the aihub old adult freq speed pause changed dataset. It achieves the following results on the evaluation set:

—Loss: 0.1846
—Cer: 4.7580

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: 8
—evalbatchsize: 8
—seed: 42
—gradientaccumulationsteps: 2
—totaltrainbatch_size: 16
—optimizer: Use adamwtorch with betas=(0.9,0.999) and epsilon=1e-08 and optimizerargs=No additional optimizer arguments
—lrschedulertype: linear
—lrschedulerwarmup_steps: 50
—num_epochs: 2
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.2852	0.1289	100	0.2569	6.5672
0.1641	0.2579	200	0.2279	6.2265
0.1599	0.3868	300	0.2275	6.3440
0.1446	0.5158	400	0.2165	6.0268
0.1147	0.6447	500	0.2075	5.9563
0.1144	0.7737	600	0.2032	5.6861
0.115	0.9026	700	0.1992	5.6097
0.0555	1.0309	800	0.1890	5.1163
0.0483	1.1599	900	0.1927	5.0752
0.0473	1.2888	1000	0.1923	4.9577
0.0418	1.4178	1100	0.1908	4.9107
0.041	1.5467	1200	0.1871	4.8578
0.0423	1.6757	1300	0.1866	4.9518
0.0388	1.8046	1400	0.1849	4.7874
0.0422	1.9336	1500	0.1846	4.7580

Framework versions

—Transformers 4.47.0.dev0
—Pytorch 2.5.0+cu121
—Datasets 3.1.0
—Tokenizers 0.20.3