ai-light-dance_singing5_ft_wav2vec2-large-xlsr-53-5gram-v4-2-1

This model is a fine-tuned version of gary109/ai-light-dance_singing4_ft_wav2vec2-large-xlsr-53-5gram-v4-2-1 on the GARY109/AI_LIGHT_DANCE - ONSET-SINGING5 dataset. It achieves the following results on the evaluation set:

• Loss: 0.1732
• Wer: 0.0831

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: 4e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 50.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4351	1.0	100	0.1948	0.0903
0.4381	2.0	200	0.1961	0.0930
0.441	3.0	300	0.1948	0.0957
0.453	4.0	400	0.1971	0.0905
0.4324	5.0	500	0.1823	0.0879
0.4561	6.0	600	0.1934	0.0893
0.4231	7.0	700	0.2088	0.0977
0.4339	8.0	800	0.1924	0.0856
0.4195	9.0	900	0.1835	0.0846
0.4162	10.0	1000	0.1869	0.0908
0.411	11.0	1100	0.1966	0.0950
0.4034	12.0	1200	0.1890	0.0879
0.4155	13.0	1300	0.1844	0.0915
0.4123	14.0	1400	0.1849	0.0891
0.4002	15.0	1500	0.1901	0.0902
0.3983	16.0	1600	0.1879	0.0865
0.3907	17.0	1700	0.1863	0.0856
0.3969	18.0	1800	0.1773	0.0836
0.3721	19.0	1900	0.1834	0.0890
0.3987	20.0	2000	0.1817	0.0852
0.3863	21.0	2100	0.1898	0.0914
0.4052	22.0	2200	0.1882	0.0857
0.3811	23.0	2300	0.1874	0.0856
0.3791	24.0	2400	0.1932	0.0885
0.3919	25.0	2500	0.1847	0.0815
0.3891	26.0	2600	0.1850	0.0852
0.3719	27.0	2700	0.1774	0.0820
0.3791	28.0	2800	0.1756	0.0825
0.3537	29.0	2900	0.1797	0.0844
0.361	30.0	3000	0.1818	0.0834
0.3619	31.0	3100	0.1747	0.0838
0.3626	32.0	3200	0.1773	0.0844
0.3632	33.0	3300	0.1775	0.0825
0.3666	34.0	3400	0.1835	0.0859
0.3581	35.0	3500	0.1859	0.0868
0.3665	36.0	3600	0.1741	0.0849
0.3495	37.0	3700	0.1790	0.0837
0.3509	38.0	3800	0.1782	0.0841
0.3621	39.0	3900	0.1759	0.0841
0.3415	40.0	4000	0.1796	0.0851
0.3508	41.0	4100	0.1777	0.0821
0.3493	42.0	4200	0.1758	0.0829
0.359	43.0	4300	0.1788	0.0848
0.3438	44.0	4400	0.1782	0.0836
0.3642	45.0	4500	0.1732	0.0831
0.3456	46.0	4600	0.1768	0.0823
0.3532	47.0	4700	0.1735	0.0834
0.3448	48.0	4800	0.1755	0.0827
0.3487	49.0	4900	0.1767	0.0833
0.3427	50.0	5000	0.1774	0.0836

Framework versions

• Transformers 4.21.0.dev0
• Pytorch 1.9.1+cu102
• Datasets 2.3.3.dev0
• Tokenizers 0.12.1