dino_finetuned_giant_10_layers_thawed

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

• Loss: 1.0387
• Accuracy: 0.7361

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: 54
• eval_batch_size: 54
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 216
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.3778	0.3145	25	2.5161	0.3373
2.0106	0.6289	50	1.8874	0.4838
1.9	0.9434	75	1.6407	0.5441
1.2966	1.2579	100	1.4907	0.5930
1.3413	1.5723	125	1.3532	0.6358
1.2871	1.8868	150	1.2731	0.6547
0.7792	2.2013	175	1.1967	0.6875
0.7153	2.5157	200	1.1761	0.6966
0.7544	2.8302	225	1.1136	0.7096
0.465	3.1447	250	1.0962	0.7187
0.414	3.4591	275	1.0997	0.7274
0.4749	3.7736	300	1.0717	0.7291
0.4742	4.0881	325	1.0425	0.7323
0.3448	4.4025	350	1.0402	0.7392
0.3341	4.7170	375	1.0387	0.7361

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1