learning_rate: 0.0003
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
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: 100
num_epochs: 10
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Cer
12.0516	0.4329	100	2.8632	1.0
3.8314	0.8658	200	1.2018	0.3760
2.2137	1.2987	300	0.8982	0.2561
1.8036	1.7316	400	0.8379	0.2249
1.5853	2.1645	500	0.7448	0.2232
1.2507	2.5974	600	0.7016	0.2235
1.3181	3.0303	700	0.6789	0.2213
0.9621	3.4632	800	0.6693	0.2024
0.9775	3.8961	900	0.6462	0.2223
0.8169	4.3290	1000	0.6833	0.1881
0.7636	4.7619	1100	0.7182	0.1868
0.7888	5.1948	1200	0.6615	0.1914
0.8115	5.6277	1300	0.6621	0.2100
0.8981	6.0606	1400	0.6533	0.2102
0.8507	6.4935	1500	0.6728	0.2085
0.8168	6.9264	1600	0.6478	0.2064
0.7990	7.3593	1700	0.6402	0.2051
0.7404	7.7922	1800	0.6456	0.2015
0.7587	8.2251	1900	0.6457	0.1992
0.7361	8.6580	2000	0.6489	0.1986
0.7172	9.0909	2100	0.6515	0.1988
0.7141	9.5238	2200	0.6518	0.1980
0.7211	9.9567	2300	0.6499	0.1990

ctaguchi/hubert-xlarge-ll60k-gui

hubert-xlarge-ll60k-gui