wav2vec2-2

This model is a fine-tuned version of jadasdn/wav2vec2-1 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.8353
• Wer: 0.3593

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: 8
• 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: 1000
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4516	0.5	500	0.4973	0.3748
0.4624	1.0	1000	0.4486	0.3958
0.5211	1.5	1500	0.5173	0.3916
0.5317	2.0	2000	0.4713	0.3992
0.4277	2.5	2500	0.4859	0.3888
0.4495	3.0	3000	0.4962	0.3862
0.3712	3.5	3500	0.5237	0.3899
0.3855	4.0	4000	0.4975	0.3850
0.3254	4.5	4500	0.5405	0.3897
0.331	5.0	5000	0.5255	0.3950
0.2907	5.5	5500	0.5646	0.3852
0.2949	6.0	6000	0.5782	0.3965
0.2521	6.5	6500	0.5563	0.3879
0.2663	7.0	7000	0.5627	0.3829
0.2342	7.5	7500	0.6145	0.3872
0.2374	8.0	8000	0.5860	0.3883
0.2099	8.5	8500	0.6920	0.3810
0.2133	9.0	9000	0.6354	0.3895
0.1887	9.5	9500	0.6618	0.3813
0.1924	10.0	10000	0.6522	0.3850
0.1728	10.5	10500	0.6324	0.3813
0.1797	11.0	11000	0.6637	0.3882
0.163	11.5	11500	0.6806	0.3799
0.1623	12.0	12000	0.6801	0.3811
0.149	12.5	12500	0.6723	0.3832
0.1493	13.0	13000	0.7032	0.3888
0.1389	13.5	13500	0.7294	0.3793
0.1383	14.0	14000	0.7311	0.3800
0.127	14.5	14500	0.7088	0.3773
0.127	15.0	15000	0.7352	0.3775
0.1159	15.5	15500	0.7886	0.3792
0.114	16.0	16000	0.7582	0.3802
0.1103	16.5	16500	0.7662	0.3717
0.1088	17.0	17000	0.7855	0.3704
0.1021	17.5	17500	0.7326	0.3717
0.104	18.0	18000	0.7518	0.3723
0.096	18.5	18500	0.7468	0.3743
0.0914	19.0	19000	0.7906	0.3741
0.0881	19.5	19500	0.7879	0.3740
0.0908	20.0	20000	0.8111	0.3676
0.0832	20.5	20500	0.8114	0.3681
0.0848	21.0	21000	0.8178	0.3651
0.0762	21.5	21500	0.8212	0.3686
0.0728	22.0	22000	0.8142	0.3673
0.074	22.5	22500	0.8177	0.3666
0.0691	23.0	23000	0.8323	0.3662
0.0689	23.5	23500	0.8020	0.3678
0.0643	24.0	24000	0.8145	0.3653
0.0647	24.5	24500	0.8376	0.3594
0.0654	25.0	25000	0.8307	0.3608
0.061	25.5	25500	0.8432	0.3600
0.0573	26.0	26000	0.8361	0.3629
0.0583	26.5	26500	0.8363	0.3625
0.054	27.0	27000	0.8277	0.3625
0.058	27.5	27500	0.8354	0.3614
0.0531	28.0	28000	0.8363	0.3595
0.0522	28.5	28500	0.8429	0.3588
0.0503	29.0	29000	0.8267	0.3595
0.0504	29.5	29500	0.8401	0.3597
0.0511	30.0	30000	0.8353	0.3593

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0