dinov2-base-ODIR-5K

This model is a fine-tuned version of facebook/dinov2-base on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.5700
• Accuracy: 0.7189
• F1: 0.6333

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-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 6

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.6374	0.9858	52	0.6186	0.6778	0.2031
0.5789	1.9905	105	0.5661	0.7153	0.3794
0.5368	2.9953	158	0.5334	0.7407	0.5756
0.4162	4.0	211	0.5747	0.6983	0.6198
0.3679	4.9858	263	0.5700	0.7189	0.6333
0.2431	5.9147	312	0.6111	0.7564	0.6331

Framework versions

• Transformers 4.40.2
• Pytorch 2.2.1+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1