wav2vec2-base-timit-demo-google-colab

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

• Loss: 0.5406
• Wer: 0.3393

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
3.5789	1.0	500	2.1382	1.0004
0.9156	2.01	1000	0.5722	0.5177
0.446	3.01	1500	0.4804	0.4662
0.3014	4.02	2000	0.4326	0.4278
0.2326	5.02	2500	0.4359	0.4074
0.1928	6.02	3000	0.4353	0.4004
0.16	7.03	3500	0.4414	0.3838
0.1421	8.03	4000	0.4251	0.3865
0.1251	9.04	4500	0.5128	0.3786
0.1085	10.04	5000	0.4985	0.3841
0.1005	11.04	5500	0.5047	0.3750
0.0876	12.05	6000	0.4934	0.3651
0.0806	13.05	6500	0.5299	0.3684
0.0723	14.06	7000	0.5206	0.3639
0.0682	15.06	7500	0.5409	0.3623
0.0637	16.06	8000	0.5135	0.3657
0.0573	17.07	8500	0.5416	0.3564
0.0563	18.07	9000	0.5378	0.3610
0.046	19.08	9500	0.5530	0.3523
0.0475	20.08	10000	0.5603	0.3544
0.0421	21.08	10500	0.5027	0.3532
0.0354	22.09	11000	0.5579	0.3511
0.0347	23.09	11500	0.5536	0.3590
0.0326	24.1	12000	0.5594	0.3504
0.0277	25.1	12500	0.5058	0.3470
0.0261	26.1	13000	0.5304	0.3488
0.0255	27.11	13500	0.5499	0.3407
0.0243	28.11	14000	0.5489	0.3415
0.0238	29.12	14500	0.5406	0.3393

Framework versions

• Transformers 4.24.0
• Pytorch 2.0.1+cu118
• Datasets 1.18.3
• Tokenizers 0.13.3