vit-base-patch16-224-in21k-lora

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3377
• Accuracy: 0.9013
• Pca Pca Loss: 0.7490
• Pca Pca Accuracy: 0.8261

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: 0.002
• train_batch_size: 128
• eval_batch_size: 128
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 512
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Pca Loss	Pca Accuracy
1.0586	0.9923	97	0.5664	0.8731	1.1687	0.8079
0.8991	1.9949	195	0.4300	0.8861	0.9371	0.8158
0.8897	2.9974	293	0.3968	0.8924	0.8579	0.8205
0.8447	4.0	391	0.3713	0.8974	0.8137	0.823
0.6856	4.9923	488	0.3585	0.8964	0.7872	0.8236
0.8925	5.9949	586	0.3513	0.8998	0.7704	0.8252
0.8224	6.9974	684	0.3455	0.9009	0.7607	0.8242
0.8563	8.0	782	0.3417	0.9019	0.7535	0.8257
0.8198	8.9923	879	0.3388	0.9023	0.7501	0.827
0.7705	9.9233	970	0.3377	0.9013	0.7490	0.8261

Framework versions

• PEFT 0.13.0
• Transformers 4.45.1
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.0