vit-base-patch16-224-in21k-lora

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on food101 dataset. It achieves the following results on the evaluation set:

• trainable params: 667,493 || all params: 86,543,818 || trainable%: 0.7713
• Loss: 0.3400

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.005
• train_batch_size: 256
• eval_batch_size: 256
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 1024
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.8448	1.0	74	0.4689
0.7281	2.0	148	0.4009
0.6533	3.0	222	0.3697
0.5799	4.0	296	0.3520
0.5547	5.0	370	0.3400

Framework versions

• PEFT 0.11.1
• Transformers 4.43.4
• Pytorch 2.2.1+cu118
• Datasets 2.21.0
• Tokenizers 0.19.1