Official pytorch code for "APP: Anytime Progressive Pruning"

Last update: Nov 22, 2022

Overview

APP: Anytime Progressive Pruning

Diganta Misra^1,2,3, Bharat Runwal^2,4, Tianlong Chen⁵, Zhangyang Wang⁵, Irina Rish^1,3

¹ Mila - Quebec AI Institute,² Landskape AI,³ UdeM,⁴ IIT-Delhi,⁵ VITA, UT-Austin

Requirements

To create a new conda environment with the dependencies used in this project, do:

conda env create -f app.yml

For running the code on Restricted-Imagenet Dataset, first install the robustness library from here and provide the imagenet_path argument as the path to the imaganet data folder.

Run the Code

Here is an example of running the Anytime Progressive Pruning (APP) on Cifar-10 dataset with 8 megabatches in total:

python main_anytime_train.py \
    --data ../data \
    --dataset cifar10 \
    --arch resnet50 \
    --seed 1 \
    --epochs 50 \
    --decreasing_lr 20,40 \
    --batch_size 64 \
    --weight_decay 1e-4 \
    --meta_batch_size 6250 \
    --meta_batch_number 8 \
    --sparsity_level 4.5 \
    --snip_size 0.20 \
    --save_dir c10_r50

One-Shot pruning :

python main_anytime_one.py \
    --data ../data \
    --dataset cifar10 \
    --arch resnet50 \
    --seed 1 \
    --epochs 50 \
    --decreasing_lr 20,40 \
    --batch_size 64 \
    --weight_decay 1e-4 \
    --meta_batch_size 6250 \
    --meta_batch_number 8 \
    --sparsity_level 4.5 \
    --snip_size 0.20 \
    --save_dir c10_OSP_r18

Baseline :

python main_anytime_baseline.py \
    --data ../data \
    --dataset cifar10 \
    --arch resnet50 \
    --seed 1 \
    --epochs 50 \
    --decreasing_lr 20,40 \
    --batch_size 64 \
    --weight_decay 1e-4 \
    --meta_batch_size 6250 \
    --meta_batch_number 8 \
    --save_dir c10_BASE_r50

Cite:

@misc{misra2022app,
    title={APP: Anytime Progressive Pruning},
    author={Diganta Misra and Bharat Runwal and Tianlong Chen and Zhangyang Wang and Irina Rish},
    year={2022},
    eprint={2204.01640},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

Official pytorch code for "APP: Anytime Progressive Pruning"

Related tags

Overview

APP: Anytime Progressive Pruning

Requirements

Run the Code

Cite:

Owner

Landskape AI

Exploiting Robust Unsupervised Video Person Re-identification

Adaout is a practical and flexible regularization method with high generalization and interpretability

The most simple and minimalistic navigation dashboard.

PyTorch implementation of DirectCLR from paper Understanding Dimensional Collapse in Contrastive Self-supervised Learning

SAFL: A Self-Attention Scene Text Recognizer with Focal Loss

Image-to-Image Translation with Conditional Adversarial Networks (Pix2pix) implementation in keras

Official Implementation of "Transformers Can Do Bayesian Inference"

Pyeventbus: a publish/subscribe event bus

Implementation of Uformer, Attention-based Unet, in Pytorch

The official implementation of CVPR 2021 Paper: Improving Weakly Supervised Visual Grounding by Contrastive Knowledge Distillation.

Deformable DETR is an efficient and fast-converging end-to-end object detector.

Official implementation of the NeurIPS 2021 paper Online Learning Of Neural Computations From Sparse Temporal Feedback

BADet: Boundary-Aware 3D Object Detection from Point Clouds (Pattern Recognition 2022)

LVI-SAM: Tightly-coupled Lidar-Visual-Inertial Odometry via Smoothing and Mapping

Generating Fractals on Starknet with Cairo

Training, generation, and analysis code for Learning Particle Physics by Example: Location-Aware Generative Adversarial Networks for Physics

Exe-to-xlsm - Simple script to create VBscript of exe and inject to xlsm

FastyAPI is a Stack boilerplate optimised for heavy loads.

InsCLR: Improving Instance Retrieval with Self-Supervision

A pytorch-version implementation codes of paper: "BSN++: Complementary Boundary Regressor with Scale-Balanced Relation Modeling for Temporal Action Proposal Generation"