Jun 19, 2025 15:20 Delete README.md May 22, 2025 07:17 Update README.md 07:07 Update README.md 07:00 Update README.md 06:55 Update README.md 06:52 Upload 2 files 06:44 Upload vit_large_patch32_224.pth ===== Training vit_large_patch32_224 =====
Epoch 1/5 | Train Acc: 87.97% | Val Acc: 88.16%
Epoch 2/5 | Train Acc: 88.31% | Val Acc: 88.42%
Epoch 3/5 | Train Acc: 89.06% | Val Acc: 88.70%
Epoch 4/5 | Train Acc: 90.13% | Val Acc: 89.34%
Epoch 5/5 | Train Acc: 91.31% | Val Acc: 89.23% May 21, 2025 16:14 Upload vit_large_patch16_224.pth ===== Training vit_large_patch16_224 =====
Epoch 1/5 | Train Acc: 88.04% | Val Acc: 88.32%
Epoch 2/5 | Train Acc: 89.53% | Val Acc: 89.89%
Epoch 3/5 | Train Acc: 91.60% | Val Acc: 90.58%
Epoch 4/5 | Train Acc: 93.46% | Val Acc: 90.49%
Epoch 5/5 | Train Acc: 94.89% | Val Acc: 91.22% May 20, 2025 21:00 Upload vit_tiny_patch16_224.pth ===== Training vit_tiny_patch16_224 =====
Epoch 1/5 | Train Acc: 87.85% | Val Acc: 87.91%
Epoch 2/5 | Train Acc: 88.47% | Val Acc: 88.59%
Epoch 3/5 | Train Acc: 89.47% | Val Acc: 88.69%
Epoch 4/5 | Train Acc: 90.95% | Val Acc: 88.95%
Epoch 5/5 | Train Acc: 92.41% | Val Acc: 89.20% 18:51 Update README.md 18:48 Update README.md 18:48 Update README.md 18:32 Update README.md 18:15 Upload vit_small_patch32_224.pth ===== Training vit_small_patch32_224 =====
Epoch 1/5 | Train Acc: 87.94% | Val Acc: 88.12%
Epoch 2/5 | Train Acc: 88.36% | Val Acc: 88.20%
Epoch 3/5 | Train Acc: 89.15% | Val Acc: 88.60%
Epoch 4/5 | Train Acc: 90.41% | Val Acc: 88.95%
Epoch 5/5 | Train Acc: 91.58% | Val Acc: 88.38% 10:53 Upload vit_base_patch32_224.pth ===== Training vit_base_patch32_224 =====
Epoch 1/5 | Train Acc: 61.57% | Val Acc: 71.96%
Epoch 2/5 | Train Acc: 69.05% | Val Acc: 74.73%
Epoch 3/5 | Train Acc: 73.89% | Val Acc: 74.59%
Epoch 4/5 | Train Acc: 76.91% | Val Acc: 77.67%
Epoch 5/5 | Train Acc: 79.54% | Val Acc: 79.54% 10:51 Upload vit_base_patch16_224.pth ===== Training vit_base_patch16_224 =====
Epoch 1/5 | Train Acc: 68.58% | Val Acc: 76.79%
Epoch 2/5 | Train Acc: 81.67% | Val Acc: 80.18%
Epoch 3/5 | Train Acc: 85.97% | Val Acc: 86.66%
Epoch 4/5 | Train Acc: 88.99% | Val Acc: 85.64%
Epoch 5/5 | Train Acc: 90.65% | Val Acc: 85.36% 10:49 Upload vit_best_model_optuna.pth HYPER PARAMETER TUNING -
We achieved a low accuracy level so to tune it and refine it we decided to tune the hyper parameters such as value, batch_size, learning_rate and weight_decay
Best hyperparameters found:
{'lr': 1.0369533463569177e-05, 'batch_size': 32, 'weight_decay': 0.009006790996778456}
Accuracy: 77.82%
Model saved at: vit_best_model_optuna.pth
 10:43 Upload vit_digital_forensics_xpu.pth 10:43 initial commit 
