Edit model card

deit-base-patch16-224-FV-finetuned-memes

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

  • Loss: 0.6769
  • Accuracy: 0.8485
  • Precision: 0.8458
  • Recall: 0.8485
  • F1: 0.8464

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: 0.00012
  • train_batch_size: 64
  • eval_batch_size: 64
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 256
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 20

Training results

Training Loss Epoch Step Validation Loss Accuracy Precision Recall F1
1.2733 0.99 20 1.0893 0.5811 0.5790 0.5811 0.5293
0.7284 1.99 40 0.7351 0.7210 0.7642 0.7210 0.7271
0.4267 2.99 60 0.5202 0.7991 0.8104 0.7991 0.8033
0.2181 3.99 80 0.4605 0.8346 0.8351 0.8346 0.8334
0.1504 4.99 100 0.5281 0.8253 0.8281 0.8253 0.8266
0.1001 5.99 120 0.4945 0.8369 0.8336 0.8369 0.8347
0.0874 6.99 140 0.5902 0.8338 0.8370 0.8338 0.8348
0.0634 7.99 160 0.6088 0.8253 0.8221 0.8253 0.8234
0.0699 8.99 180 0.6210 0.8207 0.8202 0.8207 0.8186
0.0661 9.99 200 0.5675 0.8385 0.8417 0.8385 0.8393
0.0592 10.99 220 0.6550 0.8253 0.8324 0.8253 0.8275
0.0559 11.99 240 0.6400 0.8416 0.8370 0.8416 0.8387
0.0501 12.99 260 0.6726 0.8393 0.8353 0.8393 0.8350
0.0529 13.99 280 0.6285 0.8408 0.8399 0.8408 0.8401
0.0478 14.99 300 0.6423 0.8400 0.8380 0.8400 0.8384
0.0458 15.99 320 0.6632 0.8369 0.8337 0.8369 0.8348
0.048 16.99 340 0.6719 0.8423 0.8401 0.8423 0.8404
0.0417 17.99 360 0.6807 0.8423 0.8415 0.8423 0.8408
0.0461 18.99 380 0.6732 0.8454 0.8440 0.8454 0.8438
0.044 19.99 400 0.6769 0.8485 0.8458 0.8485 0.8464

Framework versions

  • Transformers 4.24.0.dev0
  • Pytorch 1.11.0+cu102
  • Datasets 2.6.1.dev0
  • Tokenizers 0.13.1
Downloads last month
0

Evaluation results