vit-Diatome / README.md
sgonzalezsilot's picture
update model card README.md
5b2fd29
|
raw
history blame
2.38 kB
metadata
license: apache-2.0
tags:
  - generated_from_trainer
metrics:
  - accuracy
model-index:
  - name: vit-Diatome
    results: []

vit-Diatome

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.2285
  • Accuracy: 0.9429

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: 16
  • eval_batch_size: 8
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 4
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Accuracy
2.2009 0.23 100 1.9938 0.6045
1.2823 0.47 200 1.2293 0.7327
0.7569 0.7 300 0.9534 0.7868
0.7428 0.93 400 0.7906 0.8078
0.4309 1.16 500 0.5759 0.8538
0.349 1.4 600 0.5070 0.8742
0.517 1.63 700 0.5048 0.8794
0.3667 1.86 800 0.5212 0.8596
0.169 2.09 900 0.4112 0.8888
0.1443 2.33 1000 0.3294 0.9109
0.1389 2.56 1100 0.3146 0.9190
0.142 2.79 1200 0.2994 0.9208
0.0921 3.02 1300 0.2620 0.9324
0.0768 3.26 1400 0.2516 0.9336
0.061 3.49 1500 0.2425 0.9388
0.0729 3.72 1600 0.2335 0.9418
0.0757 3.95 1700 0.2285 0.9429

Framework versions

  • Transformers 4.27.1
  • Pytorch 1.13.1+cu116
  • Datasets 2.10.1
  • Tokenizers 0.13.2