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: 0.4732
• Wer: 0.3300

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
3.2982	1.0	500	1.3852	0.9990
0.8067	2.01	1000	0.5318	0.5140
0.4393	3.01	1500	0.4500	0.4570
0.3007	4.02	2000	0.4259	0.4091
0.2306	5.02	2500	0.4092	0.3962
0.1845	6.02	3000	0.3949	0.3834
0.1516	7.03	3500	0.4144	0.3759
0.1347	8.03	4000	0.3958	0.3689
0.1217	9.04	4500	0.4455	0.3754
0.1039	10.04	5000	0.4228	0.3684
0.0921	11.04	5500	0.4310	0.3566
0.082	12.05	6000	0.4549	0.3617
0.078	13.05	6500	0.4535	0.3661
0.0668	14.06	7000	0.4726	0.3557
0.0648	15.06	7500	0.4414	0.3512
0.0581	16.06	8000	0.4781	0.3548
0.057	17.07	8500	0.4626	0.3588
0.0532	18.07	9000	0.5065	0.3495
0.0442	19.08	9500	0.4645	0.3390
0.0432	20.08	10000	0.4786	0.3466
0.0416	21.08	10500	0.4487	0.3425
0.0337	22.09	11000	0.4878	0.3416
0.0305	23.09	11500	0.4787	0.3413
0.0319	24.1	12000	0.4707	0.3395
0.0262	25.1	12500	0.4875	0.3345
0.0266	26.1	13000	0.4801	0.3343
0.025	27.11	13500	0.4926	0.3320
0.022	28.11	14000	0.4894	0.3313
0.0227	29.12	14500	0.4732	0.3300

Framework versions

• Transformers 4.17.0
• Pytorch 1.11.0+cu113
• Datasets 1.18.3
• Tokenizers 0.12.1