tun_wav2vec_final

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.3109
• Wer: 0.5737
• Cer: 0.1609

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: 0.0001
• train_batch_size: 10
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 80
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.2068	5.0	300	0.9499	0.6270	0.1709
0.1824	10.0	600	1.1143	0.6395	0.1740
0.1519	15.0	900	1.4216	0.6520	0.1852
0.1387	20.0	1200	1.1372	0.6176	0.1632
0.1221	25.0	1500	1.3203	0.6364	0.1694
0.1182	30.0	1800	1.3959	0.6270	0.1782
0.099	35.0	2100	1.6996	0.6176	0.1798
0.1098	40.0	2400	1.3228	0.6113	0.1713
0.0834	45.0	2700	1.2459	0.6082	0.1582
0.0801	50.0	3000	1.1573	0.5956	0.1516
0.107	55.0	3300	1.2025	0.6019	0.1640
0.0954	60.0	3600	1.2703	0.5611	0.1593
0.0581	65.0	3900	1.2382	0.5768	0.1566
0.0582	70.0	4200	1.1088	0.5799	0.1566
0.0434	75.0	4500	1.3048	0.5831	0.1597
0.0451	80.0	4800	1.3257	0.5768	0.1640
0.0383	85.0	5100	1.3002	0.5611	0.1532
0.0384	90.0	5400	1.4335	0.5768	0.1620
0.0518	95.0	5700	1.2875	0.5737	0.1570
0.0434	100.0	6000	1.3109	0.5737	0.1609

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0
• Datasets 2.21.0
• Tokenizers 0.19.1