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
2.7507	0.7732	300	1.1444	0.2588
0.9791	1.5464	600	0.9254	0.2132
0.7596	2.3196	900	0.8280	0.2019
0.6218	3.0928	1200	0.8033	0.1773
0.4943	3.8660	1500	0.8406	0.1764
0.3881	4.6392	1800	0.8503	0.1769
0.3107	5.4124	2100	0.8299	0.1746
0.2262	6.1856	2400	0.9866	0.1758
0.162	6.9588	2700	0.9897	0.1736
0.1101	7.7320	3000	1.0380	0.1697
0.0767	8.5052	3300	1.1341	0.1722
0.0545	9.2784	3600	1.1939	0.1704

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

Rajasthani-Magadhi-Marathi-w2v-bert-2.0