learning_rate: 7e-05
train_batch_size: 16
eval_batch_size: 8
seed: 42
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.98) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
training_steps: 5000
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
3.1123	16.1290	500	2.9449	1.0	1.0
1.7882	32.2581	1000	1.2114	0.9454	0.3024
0.8477	48.3871	1500	0.6383	0.5335	0.1382
0.5491	64.5161	2000	0.5775	0.4696	0.1190
0.4421	80.6452	2500	0.6003	0.4415	0.1136
0.3477	96.7742	3000	0.5878	0.4212	0.1073
0.2959	112.9032	3500	0.6053	0.4150	0.1035
0.2065	129.0323	4000	0.6154	0.4165	0.1049
0.211	145.1613	4500	0.6213	0.3838	0.0997
0.2092	161.2903	5000	0.6193	0.3885	0.1008

Rateddany/wav2vec2-xls-r-300m-oc-1h

wav2vec2-xls-r-300m-oc-1h