vit-base-VietnameseFood

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on a Vietnamese Food dataset (https://huggingface.co/datasets/TuyenTrungLe/vietnamese_food_images) with More than 17k images were on the train set, 2k5 were on the validation set, and 5k were on the test set.

It achieves the following results on the evaluation set:

• Loss: 1.2489
• Accuracy: 0.8925

Although the loss is quite high, the model predicted well with test set with 0.8639 accuracy and a loss of 0.4871

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: 0.0002
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 6
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.4936	0.1818	100	1.5493	0.6901
0.848	0.3636	200	0.9488	0.7851
0.6619	0.5455	300	0.8240	0.7865
0.6868	0.7273	400	0.6671	0.8298
0.6127	0.9091	500	0.6296	0.8296
0.4413	1.0909	600	0.6003	0.8339
0.3484	1.2727	700	0.6349	0.8153
0.3529	1.4545	800	0.5235	0.8581
0.4104	1.6364	900	0.5407	0.8512
0.3097	1.8182	1000	0.5537	0.8423
0.2527	2.0	1100	0.4871	0.8639
0.1571	2.1818	1200	0.5507	0.8587
0.2164	2.3636	1300	0.5598	0.8585
0.1875	2.5455	1400	0.5787	0.8522
0.1314	2.7273	1500	0.5262	0.8643
0.1671	2.9091	1600	0.5686	0.8587
0.0807	3.0909	1700	0.5912	0.8633
0.0989	3.2727	1800	0.6392	0.8679
0.0586	3.4545	1900	0.6587	0.8651
0.0672	3.6364	2000	0.6542	0.8758
0.0342	3.8182	2100	0.6533	0.8786
0.0484	4.0	2200	0.7314	0.8756
0.0678	4.1818	2300	0.8517	0.8788
0.075	4.3636	2400	0.9576	0.8843
0.0201	4.5455	2500	1.0758	0.8845
0.1238	4.7273	2600	1.1375	0.8871
0.0434	4.9091	2700	1.2226	0.8877
0.0493	5.0909	2800	1.1938	0.8923
0.0055	5.2727	2900	1.2594	0.8903
0.0039	5.4545	3000	1.2709	0.8887
0.0445	5.6364	3100	1.2420	0.8921
0.0347	5.8182	3200	1.2609	0.8915
0.0657	6.0	3300	1.2489	0.8925

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.2.0
• Tokenizers 0.19.1