Edit model card

deit-base-distilled-patch16-224-75-fold5

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

  • Loss: 0.2788
  • Accuracy: 0.9535

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: 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: 0.1
  • num_epochs: 100

Training results

Training Loss Epoch Step Validation Loss Accuracy
No log 1.0 2 0.8041 0.3488
No log 2.0 4 0.6042 0.7209
No log 3.0 6 0.6582 0.6977
No log 4.0 8 0.7007 0.6977
0.6827 5.0 10 0.5653 0.6977
0.6827 6.0 12 0.4345 0.8140
0.6827 7.0 14 0.4493 0.8140
0.6827 8.0 16 0.4764 0.8140
0.6827 9.0 18 0.3636 0.8372
0.4147 10.0 20 0.2769 0.8605
0.4147 11.0 22 0.3538 0.8372
0.4147 12.0 24 0.3320 0.8605
0.4147 13.0 26 0.2668 0.8372
0.4147 14.0 28 0.2823 0.8140
0.3405 15.0 30 0.4864 0.8372
0.3405 16.0 32 0.4416 0.8372
0.3405 17.0 34 0.3585 0.8140
0.3405 18.0 36 0.4769 0.8372
0.3405 19.0 38 0.5918 0.8372
0.3074 20.0 40 0.4116 0.8605
0.3074 21.0 42 0.4578 0.8372
0.3074 22.0 44 0.4294 0.8605
0.3074 23.0 46 0.5401 0.8372
0.3074 24.0 48 0.2069 0.8837
0.2155 25.0 50 0.2216 0.8837
0.2155 26.0 52 0.3837 0.8837
0.2155 27.0 54 0.2473 0.9070
0.2155 28.0 56 0.2233 0.9070
0.2155 29.0 58 0.3489 0.8837
0.1874 30.0 60 0.4635 0.8605
0.1874 31.0 62 0.2491 0.8837
0.1874 32.0 64 0.2564 0.8837
0.1874 33.0 66 0.3655 0.8837
0.1874 34.0 68 0.2498 0.9302
0.1324 35.0 70 0.2922 0.8837
0.1324 36.0 72 0.5368 0.8837
0.1324 37.0 74 0.5739 0.8837
0.1324 38.0 76 0.5049 0.8837
0.1324 39.0 78 0.5903 0.8837
0.1222 40.0 80 0.4886 0.8837
0.1222 41.0 82 0.4174 0.8837
0.1222 42.0 84 0.5429 0.8605
0.1222 43.0 86 0.6897 0.8605
0.1222 44.0 88 0.6805 0.8605
0.1008 45.0 90 0.4073 0.8837
0.1008 46.0 92 0.4161 0.8837
0.1008 47.0 94 0.6485 0.8837
0.1008 48.0 96 0.6746 0.8837
0.1008 49.0 98 0.4433 0.8837
0.117 50.0 100 0.2788 0.9535
0.117 51.0 102 0.3441 0.8837
0.117 52.0 104 0.4663 0.8605
0.117 53.0 106 0.3300 0.9070
0.117 54.0 108 0.2531 0.9535
0.0879 55.0 110 0.2261 0.9535
0.0879 56.0 112 0.3490 0.9070
0.0879 57.0 114 0.5060 0.8837
0.0879 58.0 116 0.4847 0.8837
0.0879 59.0 118 0.3604 0.8837
0.0898 60.0 120 0.3626 0.8837
0.0898 61.0 122 0.4641 0.8837
0.0898 62.0 124 0.4606 0.8837
0.0898 63.0 126 0.3419 0.8837
0.0898 64.0 128 0.3331 0.9070
0.095 65.0 130 0.3804 0.8837
0.095 66.0 132 0.3783 0.8837
0.095 67.0 134 0.3524 0.8837
0.095 68.0 136 0.4232 0.8605
0.095 69.0 138 0.4157 0.8605
0.0735 70.0 140 0.4937 0.8605
0.0735 71.0 142 0.4616 0.8605
0.0735 72.0 144 0.4388 0.8837
0.0735 73.0 146 0.3811 0.8837
0.0735 74.0 148 0.2522 0.9302
0.0855 75.0 150 0.1677 0.9535
0.0855 76.0 152 0.1653 0.9535
0.0855 77.0 154 0.2173 0.9535
0.0855 78.0 156 0.3596 0.8837
0.0855 79.0 158 0.4653 0.8605
0.077 80.0 160 0.4739 0.8837
0.077 81.0 162 0.3609 0.8837
0.077 82.0 164 0.2334 0.9535
0.077 83.0 166 0.2125 0.9535
0.077 84.0 168 0.2595 0.9535
0.0731 85.0 170 0.3702 0.8837
0.0731 86.0 172 0.4634 0.8837
0.0731 87.0 174 0.5254 0.8837
0.0731 88.0 176 0.5813 0.8837
0.0731 89.0 178 0.5628 0.8837
0.078 90.0 180 0.5479 0.8837
0.078 91.0 182 0.5216 0.8837
0.078 92.0 184 0.5098 0.8837
0.078 93.0 186 0.5244 0.8837
0.078 94.0 188 0.5472 0.8837
0.079 95.0 190 0.5590 0.8837
0.079 96.0 192 0.5589 0.8837
0.079 97.0 194 0.5455 0.8837
0.079 98.0 196 0.5322 0.8837
0.079 99.0 198 0.5215 0.8837
0.0477 100.0 200 0.5202 0.8837

Framework versions

  • Transformers 4.41.0
  • Pytorch 2.3.0+cu121
  • Datasets 2.19.1
  • Tokenizers 0.19.1
Downloads last month
2
Safetensors
Model size
85.8M params
Tensor type
F32
·
Inference API
Drag image file here or click to browse from your device
This model can be loaded on Inference API (serverless).

Finetuned from

Evaluation results