learning_rate: 1e-05
train_batch_size: 4
eval_batch_size: 4
seed: 42
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
6.0014	0.3408	500	5.5588
4.1829	0.6817	1000	5.5769
3.6451	1.0225	1500	5.5900
3.2212	1.3633	2000	5.7961
3.1675	1.7042	2500	5.8676
3.0506	2.0450	3000	5.8910
2.9579	2.3858	3500	4.5386
2.7437	2.7267	4000	4.8996
2.4375	3.0675	4500	4.4749
2.2904	3.4083	5000	4.9543
2.3382	3.7491	5500	4.4500
2.2615	4.0900	6000	4.7052
2.1767	4.4308	6500	4.4930
2.1710	4.7716	7000	5.0903
2.1930	5.1125	7500	4.7436
2.1370	5.4533	8000	4.6483
2.1168	5.7941	8500	4.3476
2.1402	6.1350	9000	4.3592
2.0962	6.4758	9500	4.4415
2.0928	6.8166	10000	4.6206
2.0644	7.1575	10500	4.3816
2.0642	7.4983	11000	4.2540
2.0738	7.8391	11500	4.4486
1.9931	8.1800	12000	4.8557
2.0052	8.5208	12500	4.5874
2.0502	8.8616	13000	4.5356
1.9721	9.2025	13500	4.5943
2.0104	9.5433	14000	4.5202
2.0033	9.8841	14500	4.5609

JoakimRasmussen/license-plate-detr-dinov3

license-plate-detr-dinov3