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: 0.5351
• Wer: 0.3384

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
3.6311	1.0	500	2.6700	1.0
1.0104	2.01	1000	0.5289	0.5277
0.4483	3.01	1500	0.4576	0.4623
0.3089	4.02	2000	0.4483	0.4255
0.2278	5.02	2500	0.4463	0.4022
0.1886	6.02	3000	0.4653	0.3938
0.1578	7.03	3500	0.4624	0.3855
0.1429	8.03	4000	0.4420	0.3854
0.1244	9.04	4500	0.4980	0.3787
0.1126	10.04	5000	0.4311	0.3785
0.1082	11.04	5500	0.5114	0.3782
0.0888	12.05	6000	0.5392	0.3725
0.0835	13.05	6500	0.6011	0.3941
0.074	14.06	7000	0.5030	0.3652
0.0667	15.06	7500	0.5041	0.3583
0.0595	16.06	8000	0.5125	0.3605
0.0578	17.07	8500	0.5206	0.3592
0.0573	18.07	9000	0.5208	0.3643
0.0469	19.08	9500	0.4670	0.3537
0.0442	20.08	10000	0.5388	0.3497
0.0417	21.08	10500	0.5213	0.3581
0.0361	22.09	11000	0.5096	0.3465
0.0338	23.09	11500	0.5178	0.3459
0.0333	24.1	12000	0.5240	0.3490
0.0256	25.1	12500	0.5438	0.3464
0.0248	26.1	13000	0.5182	0.3412
0.0231	27.11	13500	0.5628	0.3423
0.0228	28.11	14000	0.5416	0.3419
0.0223	29.12	14500	0.5351	0.3384

Framework versions

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