swin-finetuned-food101

This model is a fine-tuned version of microsoft/swin-base-patch4-window7-224 on the food101 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4401
• Accuracy: 0.9220

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: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.0579	1.0	1183	0.4190	0.9102
0.0129	2.0	2366	0.4179	0.9155
0.0076	3.0	3549	0.4219	0.9198
0.0197	4.0	4732	0.4487	0.9160
0.0104	5.0	5915	0.4414	0.9210
0.0007	6.0	7098	0.4401	0.9220
0.0021	7.0	8281	0.4401	0.9220
0.0015	8.0	9464	0.4401	0.9220
0.0056	9.0	10647	0.4401	0.9220
0.0019	10.0	11830	0.4401	0.9220

Framework versions

• Transformers 4.25.1
• Pytorch 1.13.0+cu116
• Datasets 2.7.1
• Tokenizers 0.13.2