Edit model card

vit-base-letter

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

  • Loss: 0.0515
  • Accuracy: 0.9881

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: 4
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Accuracy
0.5539 0.12 100 0.5576 0.9308
0.2688 0.25 200 0.2371 0.9665
0.1568 0.37 300 0.1829 0.9688
0.1684 0.49 400 0.1611 0.9662
0.1584 0.62 500 0.1340 0.9673
0.1569 0.74 600 0.1933 0.9531
0.0992 0.86 700 0.1031 0.9781
0.0573 0.98 800 0.1024 0.9781
0.0359 1.11 900 0.0950 0.9804
0.0961 1.23 1000 0.1200 0.9723
0.0334 1.35 1100 0.0995 0.975
0.0855 1.48 1200 0.0791 0.9815
0.0902 1.6 1300 0.0981 0.9765
0.0583 1.72 1400 0.1192 0.9712
0.0683 1.85 1500 0.0692 0.9846
0.1188 1.97 1600 0.0931 0.9785
0.0366 2.09 1700 0.0919 0.9804
0.0276 2.21 1800 0.0667 0.9846
0.0309 2.34 1900 0.0599 0.9858
0.0183 2.46 2000 0.0892 0.9769
0.0431 2.58 2100 0.0663 0.985
0.0424 2.71 2200 0.0643 0.9862
0.0453 2.83 2300 0.0646 0.9862
0.0528 2.95 2400 0.0550 0.985
0.0045 3.08 2500 0.0579 0.9846
0.007 3.2 2600 0.0517 0.9885
0.0048 3.32 2700 0.0584 0.9865
0.019 3.44 2800 0.0560 0.9873
0.0038 3.57 2900 0.0515 0.9881
0.0219 3.69 3000 0.0527 0.9881
0.0117 3.81 3100 0.0523 0.9888
0.0035 3.94 3200 0.0559 0.9865

Framework versions

  • Transformers 4.26.1
  • Pytorch 1.13.0
  • Datasets 2.1.0
  • Tokenizers 0.13.2
Downloads last month
46
Safetensors
Model size
85.8M params
Tensor type
F32
·
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Finetuned from

Dataset used to train pittawat/vit-base-letter

Space using pittawat/vit-base-letter 1