<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
image_classification_food_model
This model is a fine-tuned version of google/vit-base-patch16-224-in21k on the food101 dataset. It achieves the following results on the evaluation set:
- Loss: 1.6474
- Accuracy: 0.893
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: 3
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 2.7587 | 0.99 | 62 | 2.5481 | 0.844 |
| 1.8903 | 2.0 | 125 | 1.8096 | 0.874 |
| 1.6502 | 2.98 | 186 | 1.6474 | 0.893 |
Framework versions
- Transformers 4.30.2
- Pytorch 2.0.1+cpu
- Datasets 2.13.0
- Tokenizers 0.13.3
