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
5.2313	0.6048	300	1.0011	0.2249
1.5619	1.2097	600	0.8795	0.2012
1.2230	1.8145	900	0.8405	0.1842
0.9544	2.4194	1200	0.8044	0.1818
0.8023	3.0242	1500	0.8554	0.1849
0.5931	3.6290	1800	0.9475	0.1818
0.4534	4.2339	2100	0.9800	0.1884
0.3423	4.8387	2400	1.0565	0.1805
0.2409	5.4435	2700	1.1193	0.1811
0.1870	6.0484	3000	1.3127	0.1795
0.1232	6.6532	3300	1.3930	0.1817
0.1032	7.2581	3600	1.4935	0.1806
0.0671	7.8629	3900	1.6621	0.1788
0.0452	8.4677	4200	1.7727	0.1778
0.0335	9.0726	4500	1.8195	0.1782
0.0169	9.6774	4800	1.9085	0.1772

dhasmana/Kumaoni-Bhojpuri-Chhattisgarhi-Angika-Maithili-w2v-bert-2.0

Kumaoni-Bhojpuri-Chhattisgarhi-Angika-Maithili-w2v-bert-2.0