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: 0.5449
• Wer: 0.3386

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
3.6024	1.0	500	2.0457	0.9988
0.8954	2.01	1000	0.5245	0.5466
0.4358	3.01	1500	0.4273	0.4534
0.2969	4.02	2000	0.3994	0.4168
0.2326	5.02	2500	0.3941	0.4071
0.1881	6.02	3000	0.3888	0.3898
0.1582	7.03	3500	0.4507	0.3864
0.1373	8.03	4000	0.4533	0.3994
0.1194	9.04	4500	0.4614	0.3859
0.1134	10.04	5000	0.4481	0.3877
0.0959	11.04	5500	0.4601	0.3731
0.0918	12.05	6000	0.4525	0.3699
0.083	13.05	6500	0.4994	0.3716
0.0736	14.06	7000	0.5001	0.3797
0.0648	15.06	7500	0.5118	0.3688
0.0629	16.06	8000	0.5198	0.3611
0.0556	17.07	8500	0.4928	0.3688
0.0569	18.07	9000	0.5086	0.3520
0.0476	19.08	9500	0.5250	0.3618
0.0457	20.08	10000	0.5150	0.3586
0.0396	21.08	10500	0.4951	0.3485
0.0369	22.09	11000	0.5493	0.3514
0.0338	23.09	11500	0.5507	0.3470
0.0332	24.1	12000	0.5273	0.3466
0.0294	25.1	12500	0.5267	0.3504
0.0248	26.1	13000	0.5437	0.3422
0.0268	27.11	13500	0.5236	0.3421
0.0247	28.11	14000	0.5221	0.3377
0.0177	29.12	14500	0.5449	0.3386

Framework versions

• Transformers 4.38.2
• Pytorch 2.1.0+cu121
• Datasets 1.18.3
• Tokenizers 0.15.2