Model description

More information needed

Intended uses & limitations

More information needed

More information needed

The following hyperparameters were used during training:

learning_rate: 0.0001
train_batch_size: 16
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: 2000
num_epochs: 16
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
5.8655	1.0347	2000	0.2428	0.1690
0.4234	2.0693	4000	0.2107	0.1622
0.3572	3.1040	6000	0.1980	0.1576
0.3216	4.1386	8000	0.1975	0.1493
0.2937	5.1733	10000	0.2160	0.1501
0.2708	6.2080	12000	0.1929	0.1422
0.2479	7.2426	14000	0.1928	0.1400
0.2302	8.2773	16000	0.2002	0.1348
0.217	9.3120	18000	0.1923	0.1340
0.2013	10.3466	20000	0.1928	0.1332
0.1884	11.3813	22000	0.1901	0.1262
0.1772	12.4159	24000	0.1916	0.1242
0.1671	13.4506	26000	0.1879	0.1207
0.1612	14.4853	28000	0.1877	0.1197
0.1546	15.5199	30000	0.1874	0.1190