Activation Pruning Experiments

This repository contains code for activation pruning experiments on both text and vision models.
The goal of these experiments is to study model sparsity, free energy, and phase transition behavior when pruning activations.

---

Repository Structure

activation-pruning/
│
├── text_models/
│   ├── AG_models/            # Models tested on AG News dataset
│   │   ├── LSTM/
│   │   ├── BERT/
│   │   ├── GPT2/
│   │   ├── ELECTRA/
│   │   └── T5/
│   │
│   └── Yelp_models/          # Models tested on Yelp dataset
│       ├── LSTM/
│       ├── BERT/
│       ├── GPT2/
│       ├── ELECTRA/
│       └── T5/
│
└── vision_models/
    ├── CIFAR10_models/       # Models tested on CIFAR-10 dataset
    │   ├── CNN/
    │   ├── MLP/
    │   ├── MobileNet/
    │   ├── ResNet/
    │   └── ViT/
    │
    └── Oxford_models/        # Models tested on Oxford Flowers-102 dataset
        ├── CNN/
        ├── MLP/
        ├── MobileNet/
        ├── ResNet/
        └── ViT/