Edit model card

deit-base-distilled-patch16-224-75-fold4

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: 0.2564
  • Accuracy: 0.9302

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 0.5613 0.6744
No log 2.0 4 0.6233 0.6977
No log 3.0 6 0.7506 0.6977
No log 4.0 8 0.6462 0.6977
0.5231 5.0 10 0.3889 0.8140
0.5231 6.0 12 0.3485 0.8372
0.5231 7.0 14 0.4327 0.8372
0.5231 8.0 16 0.4470 0.8372
0.5231 9.0 18 0.3067 0.8605
0.3699 10.0 20 0.3075 0.8605
0.3699 11.0 22 0.5433 0.8372
0.3699 12.0 24 0.5849 0.8140
0.3699 13.0 26 0.4509 0.8372
0.3699 14.0 28 0.5410 0.8372
0.2951 15.0 30 0.4710 0.8140
0.2951 16.0 32 0.3738 0.8605
0.2951 17.0 34 0.3704 0.8372
0.2951 18.0 36 0.2682 0.9070
0.2951 19.0 38 0.4187 0.8372
0.2279 20.0 40 0.3961 0.8372
0.2279 21.0 42 0.2976 0.8837
0.2279 22.0 44 0.5655 0.8140
0.2279 23.0 46 0.4931 0.8372
0.2279 24.0 48 0.2672 0.9070
0.1823 25.0 50 0.2795 0.9070
0.1823 26.0 52 0.3885 0.8605
0.1823 27.0 54 0.2779 0.9070
0.1823 28.0 56 0.2990 0.8605
0.1823 29.0 58 0.3090 0.8605
0.1314 30.0 60 0.2564 0.9302
0.1314 31.0 62 0.2825 0.8605
0.1314 32.0 64 0.3655 0.8605
0.1314 33.0 66 0.2486 0.9302
0.1314 34.0 68 0.2734 0.9070
0.1387 35.0 70 0.5242 0.8372
0.1387 36.0 72 0.5958 0.8372
0.1387 37.0 74 0.3472 0.8837
0.1387 38.0 76 0.3057 0.9070
0.1387 39.0 78 0.4018 0.8372
0.1266 40.0 80 0.3940 0.8372
0.1266 41.0 82 0.3568 0.8605
0.1266 42.0 84 0.3533 0.8837
0.1266 43.0 86 0.3451 0.8837
0.1266 44.0 88 0.3478 0.9070
0.1004 45.0 90 0.3195 0.9302
0.1004 46.0 92 0.3926 0.9070
0.1004 47.0 94 0.4169 0.8837
0.1004 48.0 96 0.4274 0.8837
0.1004 49.0 98 0.4061 0.9070
0.1033 50.0 100 0.4277 0.8605
0.1033 51.0 102 0.3977 0.9070
0.1033 52.0 104 0.4428 0.8605
0.1033 53.0 106 0.6753 0.8140
0.1033 54.0 108 0.6912 0.8140
0.0827 55.0 110 0.4201 0.8605
0.0827 56.0 112 0.3086 0.9302
0.0827 57.0 114 0.3150 0.9302
0.0827 58.0 116 0.4757 0.8605
0.0827 59.0 118 0.6409 0.8605
0.0889 60.0 120 0.5430 0.8837
0.0889 61.0 122 0.4044 0.8837
0.0889 62.0 124 0.3473 0.9302
0.0889 63.0 126 0.3485 0.9302
0.0889 64.0 128 0.3711 0.9302
0.088 65.0 130 0.4405 0.8837
0.088 66.0 132 0.6526 0.8605
0.088 67.0 134 0.7019 0.8605
0.088 68.0 136 0.5408 0.8605
0.088 69.0 138 0.4057 0.9302
0.0734 70.0 140 0.3797 0.9302
0.0734 71.0 142 0.3811 0.9302
0.0734 72.0 144 0.4040 0.9302
0.0734 73.0 146 0.4567 0.8837
0.0734 74.0 148 0.5161 0.9070
0.0721 75.0 150 0.5240 0.8837
0.0721 76.0 152 0.5048 0.9070
0.0721 77.0 154 0.4635 0.9070
0.0721 78.0 156 0.4510 0.9070
0.0721 79.0 158 0.4931 0.9070
0.0592 80.0 160 0.5368 0.8837
0.0592 81.0 162 0.5297 0.8837
0.0592 82.0 164 0.4722 0.9070
0.0592 83.0 166 0.4179 0.9302
0.0592 84.0 168 0.4045 0.9302
0.0634 85.0 170 0.4200 0.9302
0.0634 86.0 172 0.4497 0.9302
0.0634 87.0 174 0.4796 0.9070
0.0634 88.0 176 0.4997 0.8837
0.0634 89.0 178 0.4867 0.8837
0.0758 90.0 180 0.4478 0.9302
0.0758 91.0 182 0.4145 0.9302
0.0758 92.0 184 0.4036 0.9302
0.0758 93.0 186 0.3877 0.9302
0.0758 94.0 188 0.3747 0.9070
0.0529 95.0 190 0.3685 0.9070
0.0529 96.0 192 0.3644 0.9070
0.0529 97.0 194 0.3648 0.9070
0.0529 98.0 196 0.3663 0.9070
0.0529 99.0 198 0.3686 0.9070
0.0594 100.0 200 0.3693 0.9070

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