vit-base-patch16-224-in21k-finetuned-eurosat

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

  • Loss: 0.1770
  • Accuracy: 0.9361

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
0.687 0.04 10 0.6778 0.6026
0.6605 0.09 20 0.6359 0.7564
0.6074 0.13 30 0.5734 0.7786
0.5464 0.17 40 0.4877 0.8267
0.4606 0.21 50 0.3836 0.8914
0.379 0.26 60 0.3269 0.8877
0.2746 0.3 70 0.2403 0.9198
0.2974 0.34 80 0.2931 0.8890
0.2459 0.39 90 0.2596 0.9016
0.2507 0.43 100 0.2366 0.9123
0.2627 0.47 110 0.2084 0.9224
0.2481 0.51 120 0.2050 0.9270
0.2372 0.56 130 0.2077 0.9267
0.2468 0.6 140 0.2111 0.9230
0.2272 0.64 150 0.1964 0.9267
0.2568 0.68 160 0.1975 0.9270
0.2608 0.73 170 0.2485 0.9048
0.2641 0.77 180 0.2143 0.9227
0.2347 0.81 190 0.1921 0.9307
0.2231 0.86 200 0.1882 0.9315
0.2147 0.9 210 0.1865 0.9329
0.2028 0.94 220 0.1901 0.9294
0.1792 0.98 230 0.1868 0.9297
0.2471 1.03 240 0.2104 0.9190
0.1896 1.07 250 0.1840 0.9321
0.2181 1.11 260 0.1800 0.9318
0.1861 1.16 270 0.1815 0.9305
0.1761 1.2 280 0.1886 0.9299
0.1703 1.24 290 0.1802 0.9315
0.184 1.28 300 0.1845 0.9321
0.1864 1.33 310 0.1791 0.9342
0.1857 1.37 320 0.1760 0.9347
0.1558 1.41 330 0.1798 0.9318
0.1852 1.45 340 0.1810 0.9323
0.183 1.5 350 0.1775 0.9321
0.2055 1.54 360 0.1789 0.9337
0.207 1.58 370 0.2082 0.9208
0.2264 1.63 380 0.1733 0.9339
0.1954 1.67 390 0.1772 0.9337
0.1676 1.71 400 0.1840 0.9302
0.1727 1.75 410 0.1784 0.9305
0.204 1.8 420 0.1731 0.9353
0.1805 1.84 430 0.1805 0.9310
0.1732 1.88 440 0.1773 0.9337
0.1831 1.93 450 0.1768 0.9337
0.1906 1.97 460 0.1967 0.9259
0.1785 2.01 470 0.1765 0.9331
0.1566 2.05 480 0.1749 0.9361
0.1612 2.1 490 0.1718 0.9342
0.1504 2.14 500 0.1770 0.9361
0.1704 2.18 510 0.1721 0.9363
0.1597 2.22 520 0.1711 0.9345
0.1283 2.27 530 0.1775 0.9361
0.1697 2.31 540 0.1722 0.9361
0.1541 2.35 550 0.1729 0.9366
0.1466 2.4 560 0.1708 0.9369
0.1604 2.44 570 0.1720 0.9371
0.1798 2.48 580 0.1718 0.9382
0.134 2.52 590 0.1733 0.9371
0.1215 2.57 600 0.1749 0.9369
0.1284 2.61 610 0.1760 0.9358
0.1449 2.65 620 0.1745 0.9361
0.214 2.7 630 0.1729 0.9382
0.1684 2.74 640 0.1724 0.9369
0.143 2.78 650 0.1737 0.9377
0.1491 2.82 660 0.1753 0.9366
0.1636 2.87 670 0.1743 0.9371
0.1672 2.91 680 0.1724 0.9377
0.1501 2.95 690 0.1720 0.9374

Framework versions

  • Transformers 4.35.0
  • Pytorch 1.12.1+cu116
  • Datasets 2.4.0
  • Tokenizers 0.14.1
Downloads last month
4
Safetensors
Model size
85.8M params
Tensor type
F32
·
Inference Providers NEW
This model is not currently available via any of the supported third-party Inference Providers, and the model is not deployed on the HF Inference API.

Model tree for moreover18/vit-base-patch16-224-in21k-finetuned-eurosat

Finetuned
(1835)
this model
Finetunes
1 model

Evaluation results