README.md

---
base_model: google/vit-base-patch16-224
library_name: transformers
pipeline_tag: image-classification
tags:
  - probex
  - model-j
  - weight-space-learning
---

# Model-J: SupViT Model (model_idx_0939)

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

<p align="center">
    🌐 <a href="https://horwitz.ai/probex" target="_blank">Project</a> | 📃 <a href="https://arxiv.org/abs/2410.13569" target="_blank">Paper</a> | 💻 <a href="https://github.com/eliahuhorwitz/ProbeX" target="_blank">GitHub</a> | 🤗 <a href="https://huggingface.co/ProbeX" target="_blank">Dataset</a>
</p>

![ProbeX](https://raw.githubusercontent.com/eliahuhorwitz/ProbeX/main/imgs/poster.png)

## Model Details

| Attribute | Value |
|---|---|
| **Subset** | SupViT |
| **Split** | val |
| **Base Model** | `google/vit-base-patch16-224` |
| **Dataset** | CIFAR100 (50 classes) |

## Training Hyperparameters

| Parameter | Value |
|---|---|
| Learning Rate | 9e-05 |
| LR Scheduler | constant |
| Epochs | 5 |
| Max Train Steps | 1665 |
| Batch Size | 64 |
| Weight Decay | 0.01 |
| Seed | 939 |
| Random Crop | True |
| Random Flip | False |

## Performance

| Metric | Value |
|---|---|
| Train Accuracy | 0.9865 |
| Val Accuracy | 0.9451 |
| Test Accuracy | 0.9408 |

## Training Categories

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

`keyboard`, `can`, `leopard`, `apple`, `spider`, `porcupine`, `bed`, `beaver`, `rocket`, `plain`, `mouse`, `lizard`, `telephone`, `seal`, `fox`, `skyscraper`, `shark`, `plate`, `tiger`, `clock`, `house`, `crocodile`, `lamp`, `streetcar`, `palm_tree`, `elephant`, `man`, `orange`, `wardrobe`, `poppy`, `road`, `girl`, `caterpillar`, `tractor`, `hamster`, `woman`, `tank`, `cattle`, `bicycle`, `butterfly`, `orchid`, `bus`, `motorcycle`, `snake`, `bridge`, `lobster`, `sea`, `snail`, `camel`, `possum`