Modelpublic

JoshuaKelleyDs/quickdraw-MobileVIT-xxs-finetune

sourceHugging Faceotherupdated 2y agoView on Hugging Face

0likes16downloads

Model Card

quickdraw-MobileViT-xxs-a

This model is a fine-tuned version of apple/mobilevit-xx-small on an unknown dataset. It achieves the following results on the evaluation set:

—Loss: 1.1512
—Accuracy: 0.7126

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: 0.0008
—trainbatchsize: 512
—evalbatchsize: 512
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: cosine
—lrschedulerwarmup_steps: 5000
—num_epochs: 8
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.8007	0.5688	5000	1.7490	0.5725
1.5116	1.1377	10000	1.5185	0.6256
1.4298	1.7065	15000	1.4384	0.6438
1.3622	2.2753	20000	1.3908	0.6547
1.332	2.8441	25000	1.3210	0.6712
1.2903	3.4130	30000	1.2758	0.6824
1.2693	3.9818	35000	1.2592	0.6864
1.2391	4.5506	40000	1.2169	0.6965
1.2078	5.1195	45000	1.1928	0.7023
1.1959	5.6883	50000	1.1779	0.7059
1.1749	6.2571	55000	1.1669	0.7083
1.1713	6.8259	60000	1.1564	0.7110
1.1573	7.3948	65000	1.1524	0.7123
1.1555	7.9636	70000	1.1512	0.7126

Framework versions

—Transformers 4.41.0
—Pytorch 2.2.1
—Datasets 2.19.1
—Tokenizers 0.19.1