learning_rate: 3e-05
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 32
optimizer: Use adamw_torch 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: 2000
num_epochs: 30
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Cer
19.1979	1.0	269	10.6767	1.0
7.8522	2.0	538	6.5871	1.0
5.1584	3.0	807	4.7595	1.0
4.2463	4.0	1076	3.8167	1.0
3.6128	5.0	1345	3.5503	0.9380
3.4602	6.0	1614	3.3786	0.8843
3.3353	7.0	1883	2.9564	0.7770
2.8821	8.0	2152	2.2163	0.5431
2.493	9.0	2421	1.7572	0.3946
2.333	10.0	2690	1.4548	0.3100
2.121	11.0	2959	1.2082	0.2376
1.9937	12.0	3228	1.0847	0.2092
1.9522	13.0	3497	1.0128	0.2028
1.8843	14.0	3766	0.9655	0.2021
1.8219	15.0	4035	0.9371	0.1922
1.7844	16.0	4304	0.9400	0.1911
1.8193	17.0	4573	0.9754	0.1949
1.8547	18.0	4842	1.0153	0.1930
1.8814	19.0	5111	1.0676	0.1964
1.9084	20.0	5380	1.0797	0.2021
2.0475	21.0	5649	1.2656	0.2215
2.3855	22.0	5918	1.4324	0.2349

duck-hug-567/xls-r-300m-gaphon-base

xls-r-300m-gaphon-base