wav2vec2-4

This model is a fine-tuned version of jadasdn/wav2vec2-3 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.8336
• Wer: 0.3531

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
0.6872	0.5	500	0.3945	0.3562
0.5186	1.0	1000	0.4087	0.3701
0.5756	1.5	1500	0.4499	0.3770
0.4795	2.0	2000	0.4292	0.3754
0.379	2.5	2500	0.4430	0.3695
0.4879	3.0	3000	0.4530	0.3749
0.4527	3.5	3500	0.5052	0.3762
0.3305	4.0	4000	0.4820	0.3693
0.2707	4.5	4500	0.5045	0.3808
0.4131	5.0	5000	0.4771	0.3700
0.3045	5.5	5500	0.5130	0.3786
0.2459	6.0	6000	0.5071	0.3687
0.2674	6.5	6500	0.5515	0.3741
0.2224	7.0	7000	0.5358	0.3740
0.2228	7.5	7500	0.5648	0.3747
0.1992	8.0	8000	0.5644	0.3704
0.2089	8.5	8500	0.6098	0.3729
0.1795	9.0	9000	0.5837	0.3707
0.1584	9.5	9500	0.6143	0.3705
0.1741	10.0	10000	0.6294	0.3774
0.1461	10.5	10500	0.6406	0.3731
0.1448	11.0	11000	0.6352	0.3733
0.1318	11.5	11500	0.6338	0.3699
0.1396	12.0	12000	0.6440	0.3692
0.1226	12.5	12500	0.7047	0.3757
0.1232	13.0	13000	0.6815	0.3675
0.1168	13.5	13500	0.6607	0.3679
0.1128	14.0	14000	0.6650	0.3678
0.1046	14.5	14500	0.6944	0.3727
0.104	15.0	15000	0.7186	0.3652
0.097	15.5	15500	0.7224	0.3664
0.1022	16.0	16000	0.6928	0.3664
0.0987	16.5	16500	0.7248	0.3684
0.0905	17.0	17000	0.7157	0.3648
0.0863	17.5	17500	0.7410	0.3635
0.0829	18.0	18000	0.7629	0.3643
0.0788	18.5	18500	0.7371	0.3632
0.0799	19.0	19000	0.7554	0.3652
0.0744	19.5	19500	0.7886	0.3638
0.076	20.0	20000	0.7376	0.3631
0.0701	20.5	20500	0.7723	0.3586
0.0709	21.0	21000	0.7964	0.3613
0.0665	21.5	21500	0.7782	0.3576
0.066	22.0	22000	0.7885	0.3589
0.0621	22.5	22500	0.7906	0.3593
0.0614	23.0	23000	0.7737	0.3575
0.0626	23.5	23500	0.7903	0.3610
0.0595	24.0	24000	0.7937	0.3582
0.0572	24.5	24500	0.8263	0.3573
0.0553	25.0	25000	0.8125	0.3549
0.0529	25.5	25500	0.8154	0.3534
0.0524	26.0	26000	0.8018	0.3526
0.0513	26.5	26500	0.8160	0.3531
0.0491	27.0	27000	0.8117	0.3524
0.0456	27.5	27500	0.8263	0.3527
0.0479	28.0	28000	0.8368	0.3529
0.0459	28.5	28500	0.8325	0.3534
0.0427	29.0	29000	0.8331	0.3534
0.0458	29.5	29500	0.8315	0.3527
0.0436	30.0	30000	0.8336	0.3531

Framework versions

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