learning_rate: 0.0001
train_batch_size: 32
eval_batch_size: 8
seed: 42
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 1000
num_epochs: 30
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
4.1195	2.0	500	2.9503	1.0
0.9135	4.0	1000	1.0396	0.5771
0.2943	6.0	1500	0.9710	0.5172
0.197	8.0	2000	0.9747	0.4996
0.1539	10.0	2500	1.1273	0.4917
0.1223	12.0	3000	1.0586	0.4788
0.101	14.0	3500	1.0533	0.4733
0.0855	16.0	4000	1.1629	0.4549
0.0746	18.0	4500	1.1766	0.4586
0.0651	20.0	5000	1.2563	0.4599
0.058	22.0	5500	1.2531	0.4566
0.052	24.0	6000	1.2663	0.4519
0.0479	26.0	6500	1.2494	0.4495
0.043	28.0	7000	1.2585	0.4452
0.0392	30.0	7500	1.2774	0.4442

colabmafari/wav2vec2-base-commonvoice-demo-colabmafari

wav2vec2-base-commonvoice-demo-colabmafari