learning_rate: 5e-05
train_batch_size: 16
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 32
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: 10
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Cer
5.6299	0.7802	300	1.0014	0.2274
1.8092	1.5592	600	0.8511	0.1945
1.4229	2.3381	900	0.8032	0.1814
1.1826	3.1170	1200	0.7585	0.1784
0.9241	3.8973	1500	0.7574	0.1715
0.7019	4.6762	1800	0.7691	0.1619
0.5208	5.4551	2100	0.8555	0.1615
0.3735	6.2341	2400	0.9440	0.1617
0.2511	7.0130	2700	1.0597	0.1606
0.1586	7.7932	3000	1.0751	0.1641
0.1038	8.5722	3300	1.2033	0.1617
0.0679	9.3511	3600	1.3354	0.1608

dhasmana/Chhattisgarhi-Magadhi-Marathi-w2v-bert-2.0

Chhattisgarhi-Magadhi-Marathi-w2v-bert-2.0