learning_rate: 0.0001
train_batch_size: 32
eval_batch_size: 8
seed: 42
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 500
num_epochs: 5
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
2.4539	0.1999	500	2.3540	0.8607
2.0432	0.3998	1000	1.9838	0.7461
1.3401	0.5998	1500	1.1969	0.6512
1.0280	0.7997	2000	0.9007	0.5507
0.8373	0.9996	2500	0.7230	0.4763
0.7539	1.1995	3000	0.6302	0.4256
0.7013	1.3994	3500	0.5832	0.4064
0.6422	1.5994	4000	0.5403	0.3816
0.6222	1.7993	4500	0.5188	0.3760
0.5585	1.9992	5000	0.4848	0.3489
0.5312	2.1991	5500	0.4617	0.3359
0.4930	2.3990	6000	0.4451	0.3316
0.4727	2.5990	6500	0.4334	0.3199
0.4611	2.7989	7000	0.4141	0.3138
0.4716	2.9988	7500	0.4022	0.3080
0.3960	3.1987	8000	0.3909	0.3023
0.4068	3.3986	8500	0.3731	0.2857
0.3808	3.5986	9000	0.3665	0.2795
0.3690	3.7985	9500	0.3584	0.2767
0.3678	3.9984	10000	0.3486	0.2726
0.3349	4.1983	10500	0.3450	0.2694
0.3213	4.3982	11000	0.3408	0.2644
0.3266	4.5982	11500	0.3377	0.2609
0.3178	4.7981	12000	0.3335	0.2575
0.3116	4.9980	12500	0.3314	0.2567
0.3116	5.0	12505	0.3315	0.2566

Oreoluwa/mms-300m-yecs_langid

mms-300m-yecs_langid