learning_rate: 0.0002
train_batch_size: 1
eval_batch_size: 1
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 4
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
training_steps: 1800

Training Loss	Epoch	Step	Validation Loss
0.2889	0.3347	100	0.3401
0.2939	0.6695	200	0.3257
0.2351	1.0033	300	0.3106
0.1816	1.3381	400	0.3135
0.1476	1.6728	500	0.3209
0.1369	2.0067	600	0.3306
0.1651	2.3414	700	0.3188
0.1199	2.6762	800	0.3130
0.0569	3.0100	900	0.3447
0.0795	3.3448	1000	0.3424
0.0576	3.6795	1100	0.3551
0.0482	4.0134	1200	0.3429
0.0348	4.3481	1300	0.3598
0.0173	4.6828	1400	0.3714
0.0257	5.0167	1500	0.3725
0.0806	5.3515	1600	0.3688
0.0320	5.6862	1700	0.3779
0.0185	6.0201	1800	0.3878

AlexandreSheva/rukopys-qwen3-vl-8b-page-text-v3

rukopys-qwen3-vl-8b-page-text-v3