Modelpublic

Nachuwu/wav2vec2-fleur-mms-batch6-epoch16-finetunning

sourceHugging Facecc-by-nc-4.0updated 3y agoView on Hugging Face

0likes9downloads

Model Card

wav2vec2-fleur-mms-batch6-epoch16-finetunning

This model is a fine-tuned version of Nachuwu/wav2vec2-fleur-mms-batch6-epoch16 on the audiofolder dataset. It achieves the following results on the evaluation set:

—Loss: 0.7739
—Wer: 0.5252

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: 0.001
—trainbatchsize: 6
—evalbatchsize: 8
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 100
—num_epochs: 16
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
22.1887	0.42	100	11.1500	2.4576
16.4246	0.84	200	7.8925	1.2116
9.6917	1.27	300	5.2737	1.0255
3.615	1.69	400	3.1120	0.9835
2.7798	2.11	500	3.0797	0.9924
2.447	2.53	600	2.6319	1.0207
2.1429	2.95	700	2.3483	0.9924
1.6767	3.38	800	1.2960	0.7926
1.19	3.8	900	1.2001	0.7250
1.1818	4.22	1000	1.0052	0.5837
1.0365	4.64	1100	0.9885	0.6120
0.9911	5.06	1200	1.0018	0.5596
0.9513	5.49	1300	0.9199	0.5500
0.8994	5.91	1400	0.9272	0.5672
0.9067	6.33	1500	0.8948	0.5451
0.8964	6.75	1600	0.8798	0.5665
0.888	7.17	1700	0.8836	0.6223
0.8494	7.59	1800	0.8691	0.5830
0.8678	8.02	1900	0.8436	0.5493
0.8393	8.44	2000	0.8303	0.5431
0.8091	8.86	2100	0.8202	0.5438
0.8937	9.28	2200	0.7999	0.5341
0.7479	9.7	2300	0.8417	0.5258
0.8014	10.13	2400	0.8220	0.5369
0.8366	10.55	2500	0.8024	0.5348
0.7695	10.97	2600	0.8242	0.5686
0.761	11.39	2700	0.7992	0.5569
0.7845	11.81	2800	0.7981	0.5500
0.8056	12.24	2900	0.7900	0.5355
0.8236	12.66	3000	0.7826	0.5279
0.8135	13.08	3100	0.7834	0.5245
0.709	13.5	3200	0.7814	0.5320
0.7298	13.92	3300	0.7812	0.5258
0.7536	14.35	3400	0.7799	0.5293
0.7263	14.77	3500	0.7803	0.5224
0.7469	15.19	3600	0.7745	0.5362
0.7543	15.61	3700	0.7739	0.5252

Framework versions

—Transformers 4.36.0.dev0
—Pytorch 2.1.0+cu118
—Datasets 2.15.0
—Tokenizers 0.15.0