output_dinov3_baseline

This model is a fine-tuned version of facebook/dinov3-vitb16-pretrain-lvd1689m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0561
• Precision: 0.9575
• Recall: 0.9706
• F1: 0.9640

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-06
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Use 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: 0.1
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1
0.4260	1.0	145	0.4096	0.5780	0.6891	0.6287
0.2184	2.0	290	0.2185	0.8637	0.8519	0.8577
0.1553	3.0	435	0.1501	0.8833	0.9307	0.9064
0.1031	4.0	580	0.1093	0.9232	0.9601	0.9413
0.0828	5.0	725	0.0866	0.9308	0.9748	0.9523
0.0666	6.0	870	0.0746	0.9448	0.9716	0.9581
0.0585	7.0	1015	0.0658	0.9652	0.9622	0.9637
0.0500	8.0	1160	0.0613	0.9565	0.9695	0.9630
0.0465	9.0	1305	0.0573	0.9584	0.9685	0.9634
0.0420	10.0	1450	0.0561	0.9575	0.9706	0.9640

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.5.0
• Tokenizers 0.22.2