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
5.7058	0.7702	300	1.1669	0.2727
1.8927	1.5392	600	1.0168	0.2675
1.5253	2.3081	900	0.9380	0.2205
1.2333	3.0770	1200	0.9265	0.2168
0.9826	3.8472	1500	0.9467	0.2124
0.7729	4.6162	1800	1.0030	0.2090
0.5824	5.3851	2100	1.0575	0.2101
0.4352	6.1540	2400	1.2511	0.2121
0.2928	6.9243	2700	1.3922	0.2190
0.1889	7.6932	3000	1.5311	0.2149
0.1252	8.4621	3300	1.7120	0.2174
0.0827	9.2311	3600	1.8735	0.2156
0.0536	10.0	3900	1.9032	0.2181

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

Rajasthani-Magadhi-Chhattisgarhi-w2v-bert-2.0