Edit model card

segformer-b1-finetuned-sudoku

This model is a fine-tuned version of nvidia/mit-b1 on the mrkprc1/SudokuBoundaries2 dataset. It achieves the following results on the evaluation set:

  • Loss: 0.9826
  • Mean Iou: 0.2452
  • Mean Accuracy: 0.4999
  • Overall Accuracy: 0.4903
  • Accuracy Unlabelled: 0.9996
  • Accuracy Sudoku-boundary: 0.0001
  • Iou Unlabelled: 0.4903
  • Iou Sudoku-boundary: 0.0001

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: 6e-05
  • train_batch_size: 2
  • eval_batch_size: 2
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 50

Training results

Training Loss Epoch Step Validation Loss Mean Iou Mean Accuracy Overall Accuracy Accuracy Unlabelled Accuracy Sudoku-boundary Iou Unlabelled Iou Sudoku-boundary
0.6034 3.33 20 0.6951 0.3427 0.5173 0.5149 0.6432 0.3914 0.3940 0.2913
0.7796 6.67 40 0.7150 0.3049 0.5083 0.5021 0.8309 0.1857 0.4501 0.1597
0.4378 10.0 60 0.9772 0.2452 0.5 0.4904 1.0 0.0 0.4904 0.0
0.6804 13.33 80 1.1605 0.2452 0.5 0.4904 1.0 0.0 0.4904 0.0
0.58 16.67 100 0.9787 0.2452 0.5 0.4904 1.0 0.0 0.4904 0.0
0.6563 20.0 120 1.1860 0.2452 0.5 0.4904 1.0 0.0 0.4904 0.0
0.5128 23.33 140 0.8884 0.2457 0.5002 0.4907 0.9996 0.0009 0.4905 0.0009
0.5054 26.67 160 0.8746 0.2455 0.5002 0.4907 0.9998 0.0006 0.4905 0.0006
0.5532 30.0 180 0.9540 0.2452 0.5000 0.4905 1.0 0.0000 0.4905 0.0000
0.3238 33.33 200 0.8916 0.2470 0.5009 0.4914 0.9984 0.0035 0.4905 0.0035
0.2964 36.67 220 1.0162 0.2453 0.5000 0.4905 1.0000 0.0000 0.4905 0.0000
0.2102 40.0 240 0.9650 0.2452 0.4998 0.4903 0.9996 0.0001 0.4903 0.0001
0.623 43.33 260 0.9071 0.2461 0.5004 0.4909 0.9991 0.0017 0.4904 0.0017
0.3741 46.67 280 0.9245 0.2454 0.5000 0.4904 0.9994 0.0006 0.4903 0.0006
0.5765 50.0 300 0.9826 0.2452 0.4999 0.4903 0.9996 0.0001 0.4903 0.0001

Framework versions

  • Transformers 4.37.1
  • Pytorch 2.1.2
  • Datasets 2.16.1
  • Tokenizers 0.15.1
Downloads last month
2
Safetensors
Model size
13.7M params
Tensor type
F32
·

Finetuned from