Model-J: MAE Model (model_idx_0102)
This model is part of the Model-J dataset, introduced in:
Learning on Model Weights using Tree Experts (CVPR 2025) by Eliahu Horwitz*, Bar Cavia*, Jonathan Kahana*, Yedid Hoshen
ProbeXModel Details
| Attribute | Value |
|---|
| Subset | MAE |
| Split | test |
| Base Model | facebook/vit-mae-base |
| Dataset | CIFAR100 (50 classes) |
Training Hyperparameters
| Parameter | Value |
|---|
| Learning Rate | 0.0003 |
| LR Scheduler | constant_with_warmup |
| Epochs | 5 |
| Max Train Steps | 1665 |
| Batch Size | 64 |
| Weight Decay | 0.007 |
| Seed | 102 |
| Random Crop | False |
| Random Flip | True |
Performance
| Metric | Value |
|---|
| Train Accuracy | 0.7664 |
| Val Accuracy | 0.5509 |
| Test Accuracy | 0.5710 |
Training Categories
The model was fine-tuned on the following 50 CIFAR100 classes:
beetle, aquarium_fish, butterfly, camel, shrew, dolphin, pear, wolf, leopard, turtle, cattle, wardrobe, squirrel, crocodile, girl, plain, mountain, bowl, snail, boy, cockroach, sea, mouse, spider, plate, fox, rocket, trout, orchid, clock, flatfish, elephant, tank, kangaroo, lizard, bottle, crab, beaver, baby, apple, bicycle, motorcycle, caterpillar, castle, road, cup, bear, poppy, porcupine, maple_tree
