Model-J: ResNet Model (model_idx_0005)

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

🌐 Project | 📃 Paper | 💻 GitHub | 🤗 Dataset

Model Details

Attribute	Value
Subset	ResNet
Split	val
Base Model	`microsoft/resnet-101`
Dataset	CIFAR100 (50 classes)

Training Hyperparameters

Parameter	Value
Learning Rate	0.0003
LR Scheduler	cosine_with_restarts
Epochs	4
Max Train Steps	1332
Batch Size	64
Weight Decay	0.009
Seed	5
Random Crop	True
Random Flip	False

Performance

Metric	Value
Train Accuracy	0.9710
Val Accuracy	0.9056
Test Accuracy	0.8964

Training Categories

The model was fine-tuned on the following 50 CIFAR100 classes:

bed, cloud, mountain, streetcar, snail, palm_tree, lawn_mower, whale, shrew, caterpillar, maple_tree, tiger, telephone, butterfly, crab, rabbit, ray, mouse, bowl, leopard, cattle, camel, bridge, skyscraper, bicycle, baby, rocket, worm, house, oak_tree, snake, elephant, chair, skunk, girl, seal, motorcycle, plain, sunflower, train, chimpanzee, apple, sea, cockroach, cup, can, possum, crocodile, wardrobe, kangaroo