Edit model card

deit-base-distilled-patch16-224-75-fold2

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: 0.2625
  • 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 0.8815 0.4419
No log 2.0 4 0.6436 0.6977
No log 3.0 6 0.8488 0.6977
No log 4.0 8 0.8219 0.6977
0.6918 5.0 10 0.5491 0.6977
0.6918 6.0 12 0.4603 0.7209
0.6918 7.0 14 0.5602 0.7442
0.6918 8.0 16 0.5694 0.7442
0.6918 9.0 18 0.4430 0.8140
0.3867 10.0 20 0.3880 0.8605
0.3867 11.0 22 0.5069 0.8140
0.3867 12.0 24 0.3739 0.8605
0.3867 13.0 26 0.2981 0.8837
0.3867 14.0 28 0.3170 0.8837
0.2722 15.0 30 0.2511 0.8837
0.2722 16.0 32 0.2408 0.8837
0.2722 17.0 34 0.3751 0.8605
0.2722 18.0 36 0.3081 0.8605
0.2722 19.0 38 0.2489 0.8837
0.209 20.0 40 0.2802 0.8837
0.209 21.0 42 0.2625 0.9302
0.209 22.0 44 0.2595 0.9302
0.209 23.0 46 0.5048 0.8372
0.209 24.0 48 0.2880 0.8605
0.2027 25.0 50 0.2860 0.8372
0.2027 26.0 52 0.4067 0.8372
0.2027 27.0 54 0.2462 0.9070
0.2027 28.0 56 0.2753 0.9070
0.2027 29.0 58 0.3699 0.8140
0.1426 30.0 60 0.4983 0.8372
0.1426 31.0 62 0.3140 0.8605
0.1426 32.0 64 0.3470 0.8372
0.1426 33.0 66 0.4443 0.8372
0.1426 34.0 68 0.2583 0.8837
0.1385 35.0 70 0.2239 0.9302
0.1385 36.0 72 0.2708 0.9070
0.1385 37.0 74 0.2660 0.9070
0.1385 38.0 76 0.2754 0.9070
0.1385 39.0 78 0.4246 0.8605
0.1202 40.0 80 0.2779 0.9070
0.1202 41.0 82 0.2726 0.8837
0.1202 42.0 84 0.2536 0.9070
0.1202 43.0 86 0.2667 0.9302
0.1202 44.0 88 0.4191 0.8837
0.1211 45.0 90 0.3213 0.9302
0.1211 46.0 92 0.2290 0.9070
0.1211 47.0 94 0.3043 0.8837
0.1211 48.0 96 0.1906 0.9302
0.1211 49.0 98 0.3201 0.8605
0.1067 50.0 100 0.3062 0.8837
0.1067 51.0 102 0.2047 0.9302
0.1067 52.0 104 0.2116 0.9070
0.1067 53.0 106 0.2113 0.9302
0.1067 54.0 108 0.2340 0.9302
0.0826 55.0 110 0.2328 0.9302
0.0826 56.0 112 0.2824 0.8837
0.0826 57.0 114 0.2921 0.8837
0.0826 58.0 116 0.2608 0.9302
0.0826 59.0 118 0.2650 0.9302
0.0894 60.0 120 0.2878 0.9070
0.0894 61.0 122 0.2935 0.9070
0.0894 62.0 124 0.2656 0.9302
0.0894 63.0 126 0.3438 0.9070
0.0894 64.0 128 0.2840 0.9302
0.0964 65.0 130 0.2711 0.9070
0.0964 66.0 132 0.2888 0.9070
0.0964 67.0 134 0.2723 0.9070
0.0964 68.0 136 0.2563 0.8837
0.0964 69.0 138 0.2336 0.9302
0.0711 70.0 140 0.2386 0.9302
0.0711 71.0 142 0.2482 0.9070
0.0711 72.0 144 0.2821 0.9070
0.0711 73.0 146 0.2941 0.8837
0.0711 74.0 148 0.2564 0.9070
0.0824 75.0 150 0.2509 0.9302
0.0824 76.0 152 0.2544 0.9302
0.0824 77.0 154 0.2474 0.9302
0.0824 78.0 156 0.2375 0.9302
0.0824 79.0 158 0.2389 0.9302
0.0691 80.0 160 0.2371 0.9302
0.0691 81.0 162 0.2393 0.9302
0.0691 82.0 164 0.2523 0.9070
0.0691 83.0 166 0.2677 0.8837
0.0691 84.0 168 0.2945 0.8837
0.0638 85.0 170 0.3245 0.8605
0.0638 86.0 172 0.2960 0.8837
0.0638 87.0 174 0.2658 0.9302
0.0638 88.0 176 0.2614 0.9302
0.0638 89.0 178 0.2613 0.9302
0.0705 90.0 180 0.2549 0.9302
0.0705 91.0 182 0.2510 0.9302
0.0705 92.0 184 0.2514 0.9302
0.0705 93.0 186 0.2522 0.9302
0.0705 94.0 188 0.2504 0.9070
0.0666 95.0 190 0.2459 0.9302
0.0666 96.0 192 0.2424 0.9302
0.0666 97.0 194 0.2428 0.9302
0.0666 98.0 196 0.2451 0.9302
0.0666 99.0 198 0.2471 0.9302
0.0566 100.0 200 0.2482 0.9302

Framework versions

  • Transformers 4.41.0
  • Pytorch 2.3.0+cu121
  • Datasets 2.19.1
  • Tokenizers 0.19.1
Downloads last month
1
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