wav2vec2-3

This model is a fine-tuned version of jadasdn/wav2vec2-2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.9689
• Wer: 0.3821

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
0.5494	0.5	500	0.4696	0.3879
0.4979	1.0	1000	0.5284	0.4156
0.438	1.5	1500	0.5127	0.4121
0.4577	2.0	2000	0.5421	0.4162
0.3528	2.5	2500	0.5667	0.4117
0.3685	3.0	3000	0.5354	0.4109
0.3047	3.5	3500	0.5899	0.4244
0.3103	4.0	4000	0.5587	0.4107
0.2626	4.5	4500	0.6265	0.4255
0.2735	5.0	5000	0.6153	0.4104
0.2244	5.5	5500	0.5916	0.4058
0.2418	6.0	6000	0.5921	0.4134
0.2013	6.5	6500	0.6632	0.4079
0.211	7.0	7000	0.6563	0.4206
0.182	7.5	7500	0.6546	0.4096
0.1912	8.0	8000	0.6429	0.4043
0.1621	8.5	8500	0.7105	0.4060
0.1695	9.0	9000	0.6985	0.4027
0.1512	9.5	9500	0.7729	0.4060
0.1521	10.0	10000	0.7261	0.4047
0.1355	10.5	10500	0.7665	0.4097
0.1385	11.0	11000	0.7366	0.4044
0.1275	11.5	11500	0.7494	0.4046
0.1278	12.0	12000	0.7756	0.4028
0.1165	12.5	12500	0.8292	0.4064
0.1176	13.0	13000	0.7808	0.4053
0.1091	13.5	13500	0.8157	0.4052
0.1092	14.0	14000	0.8295	0.4071
0.1024	14.5	14500	0.8617	0.4032
0.1005	15.0	15000	0.8066	0.4042
0.0932	15.5	15500	0.8588	0.4022
0.0932	16.0	16000	0.8429	0.4036
0.087	16.5	16500	0.8975	0.4037
0.0866	17.0	17000	0.8916	0.3976
0.0835	17.5	17500	0.8566	0.3955
0.0828	18.0	18000	0.8581	0.3993
0.0756	18.5	18500	0.8827	0.3962
0.0789	19.0	19000	0.9072	0.3957
0.071	19.5	19500	0.9349	0.3920
0.0719	20.0	20000	0.9064	0.3931
0.0698	20.5	20500	0.9208	0.3932
0.0671	21.0	21000	0.8935	0.3927
0.0656	21.5	21500	0.9271	0.3936
0.0612	22.0	22000	0.9792	0.3979
0.0576	22.5	22500	0.9530	0.3917
0.0588	23.0	23000	0.9617	0.3928
0.0532	23.5	23500	0.9754	0.3854
0.0563	24.0	24000	0.9559	0.3915
0.053	24.5	24500	0.9845	0.3894
0.0512	25.0	25000	0.9629	0.3876
0.0516	25.5	25500	0.9565	0.3873
0.0491	26.0	26000	0.9726	0.3877
0.0473	26.5	26500	0.9831	0.3877
0.0441	27.0	27000	0.9645	0.3839
0.042	27.5	27500	1.0040	0.3851
0.0458	28.0	28000	0.9685	0.3839
0.043	28.5	28500	0.9687	0.3841
0.0399	29.0	29000	0.9764	0.3831
0.0421	29.5	29500	0.9691	0.3828
0.041	30.0	30000	0.9689	0.3821

Framework versions

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