Edit model card

vit-base-patch16-224-U8-40b

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

  • Loss: 0.5666
  • Accuracy: 0.8824

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: 5.5e-05
  • train_batch_size: 32
  • eval_batch_size: 32
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 128
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.05
  • num_epochs: 40

Training results

Training Loss Epoch Step Validation Loss Accuracy
1.3457 1.0 20 1.3070 0.4706
1.1498 2.0 40 1.0956 0.5686
0.8293 3.0 60 0.8270 0.6471
0.5448 4.0 80 0.6145 0.8235
0.3525 5.0 100 0.6439 0.7451
0.2436 6.0 120 0.5427 0.8235
0.195 7.0 140 0.6276 0.7843
0.1629 8.0 160 0.7868 0.7255
0.1697 9.0 180 0.8245 0.7255
0.1324 10.0 200 0.6599 0.8235
0.1714 11.0 220 0.7453 0.7647
0.0908 12.0 240 0.5666 0.8824
0.0812 13.0 260 0.9997 0.7451
0.0672 14.0 280 0.8049 0.8039
0.0843 15.0 300 0.6723 0.8431
0.0946 16.0 320 0.8892 0.7451
0.0684 17.0 340 1.1429 0.7451
0.0711 18.0 360 1.1384 0.7451
0.0677 19.0 380 1.0296 0.7843
0.0562 20.0 400 0.9803 0.7647
0.0688 21.0 420 0.9401 0.7843
0.0576 22.0 440 1.0823 0.7843
0.0892 23.0 460 1.0819 0.7255
0.063 24.0 480 1.0756 0.7647
0.055 25.0 500 0.9693 0.7647
0.0407 26.0 520 1.0132 0.7451
0.0562 27.0 540 1.0267 0.7843
0.0365 28.0 560 1.0530 0.7451
0.0363 29.0 580 0.9277 0.7843
0.0392 30.0 600 0.9798 0.8039
0.0374 31.0 620 1.0239 0.8039
0.0386 32.0 640 1.0221 0.8039
0.0345 33.0 660 1.0239 0.7843
0.035 34.0 680 1.0163 0.8039
0.0367 35.0 700 1.0902 0.8039
0.0219 36.0 720 1.1079 0.7843
0.0263 37.0 740 1.0727 0.8039
0.0261 38.0 760 1.0471 0.8039
0.0193 39.0 780 1.0347 0.8039
0.0301 40.0 800 1.0319 0.8039

Framework versions

  • Transformers 4.36.2
  • Pytorch 2.1.2+cu118
  • Datasets 2.16.1
  • Tokenizers 0.15.0
Downloads last month
132
Safetensors
Model size
85.8M params
Tensor type
F32
·

Finetuned from

Evaluation results