Add model card
Browse files
README.md
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: google/vit-base-patch16-224
|
| 3 |
+
library_name: transformers
|
| 4 |
+
pipeline_tag: image-classification
|
| 5 |
+
tags:
|
| 6 |
+
- probex
|
| 7 |
+
- model-j
|
| 8 |
+
- weight-space-learning
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# Model-J: SupViT Model (model_idx_0078)
|
| 12 |
+
|
| 13 |
+
This model is part of the **Model-J** dataset, introduced in:
|
| 14 |
+
|
| 15 |
+
**Learning on Model Weights using Tree Experts** (CVPR 2025) by Eliahu Horwitz*, Bar Cavia*, Jonathan Kahana*, Yedid Hoshen
|
| 16 |
+
|
| 17 |
+
<p align="center">
|
| 18 |
+
🌐 <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>
|
| 19 |
+
</p>
|
| 20 |
+
|
| 21 |
+
|
| 22 |
+
|
| 23 |
+
## Model Details
|
| 24 |
+
|
| 25 |
+
| Attribute | Value |
|
| 26 |
+
|---|---|
|
| 27 |
+
| **Subset** | SupViT |
|
| 28 |
+
| **Split** | train |
|
| 29 |
+
| **Base Model** | `google/vit-base-patch16-224` |
|
| 30 |
+
| **Dataset** | CIFAR100 (50 classes) |
|
| 31 |
+
|
| 32 |
+
## Training Hyperparameters
|
| 33 |
+
|
| 34 |
+
| Parameter | Value |
|
| 35 |
+
|---|---|
|
| 36 |
+
| Learning Rate | 3e-05 |
|
| 37 |
+
| LR Scheduler | constant_with_warmup |
|
| 38 |
+
| Epochs | 5 |
|
| 39 |
+
| Max Train Steps | 1665 |
|
| 40 |
+
| Batch Size | 64 |
|
| 41 |
+
| Weight Decay | 0.05 |
|
| 42 |
+
| Seed | 78 |
|
| 43 |
+
| Random Crop | True |
|
| 44 |
+
| Random Flip | False |
|
| 45 |
+
|
| 46 |
+
## Performance
|
| 47 |
+
|
| 48 |
+
| Metric | Value |
|
| 49 |
+
|---|---|
|
| 50 |
+
| Train Accuracy | 0.9871 |
|
| 51 |
+
| Val Accuracy | 0.9331 |
|
| 52 |
+
| Test Accuracy | 0.9264 |
|
| 53 |
+
|
| 54 |
+
## Training Categories
|
| 55 |
+
|
| 56 |
+
The model was fine-tuned on the following 50 CIFAR100 classes:
|
| 57 |
+
|
| 58 |
+
`girl`, `house`, `bowl`, `spider`, `bear`, `caterpillar`, `tiger`, `fox`, `television`, `camel`, `squirrel`, `plate`, `bed`, `beaver`, `mouse`, `chimpanzee`, `bridge`, `trout`, `kangaroo`, `dolphin`, `woman`, `ray`, `dinosaur`, `butterfly`, `telephone`, `orchid`, `cattle`, `maple_tree`, `orange`, `otter`, `sweet_pepper`, `oak_tree`, `lobster`, `baby`, `sea`, `bottle`, `flatfish`, `crocodile`, `leopard`, `whale`, `cockroach`, `lamp`, `clock`, `mountain`, `cloud`, `cup`, `bee`, `rabbit`, `seal`, `raccoon`
|