Model-J: MAE Model (model_idx_0979)
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 | val |
| Base Model | facebook/vit-mae-base |
| Dataset | CIFAR100 (50 classes) |
Training Hyperparameters
| Parameter | Value |
|---|
| Learning Rate | 0.0005 |
| LR Scheduler | cosine |
| Epochs | 3 |
| Max Train Steps | 999 |
| Batch Size | 64 |
| Weight Decay | 0.05 |
| Seed | 979 |
| Random Crop | False |
| Random Flip | False |
Performance
| Metric | Value |
|---|
| Train Accuracy | 0.5511 |
| Val Accuracy | 0.4613 |
| Test Accuracy | 0.4818 |
Training Categories
The model was fine-tuned on the following 50 CIFAR100 classes:
dinosaur, keyboard, rocket, cloud, snake, skyscraper, boy, plain, cup, girl, butterfly, wolf, elephant, maple_tree, sunflower, rose, camel, tractor, willow_tree, skunk, whale, pear, road, trout, clock, bridge, television, streetcar, caterpillar, kangaroo, bear, orchid, seal, cockroach, bus, orange, poppy, apple, lawn_mower, can, raccoon, mountain, pickup_truck, plate, motorcycle, lamp, otter, mouse, aquarium_fish, castle
