PARTIME: Scalable and Parallel Processing Over Time with Deep Neural Networks

This directory contains the code of the PARTIME library presented in the manuscript "PARTIME: Scalable and Parallel Processing Over Time with Deep Neural Networks".

Technical report

Requirements

The main requirements for this library are

• PyTorch >= 11 with CUDA >= 11.3
• wandb (for experimentD.py logging of accuracy over time)
• torchvision
• pandas

Content

The directory contains two subdirectories and 5 python files that help replicate the experiments described in the paper.

Directories

• extras: contains code used to generate sequential versions of the ResNet
• partime: contains the PARTIME library

Files

• common.py: contains common code used in other scripts, mainly pipeline execution and time measurement
• experimentA.py: contains the code to replicate V.A - Launch with python ./experimentA.py
• experimentB.py: contains the code to replicate V.B - Launch with python ./experimentB.py
• experimentC.py: contains the code to replicate V.C - Launch with python ./experimentC.py
• experimentD.py: contains the code to replicate V.D - Launch with python ./experimentD.py --lr=<lr> --batch_size=<batch_size> --stages=<n_stages> --max_epochs=<n_epochs>. The script contains the code to log data with wandb. If wandb is not to be used, comment the import and set DEBUG=True at line 27

Acknowledgement

This software was developed in the context of some of the activities of the PRIN 2017 project RexLearn, funded by the Italian Ministry of Education, University and Research (grant no. 2017TWNMH2).