Poros

Poros is an automatic and adaptive resource partitioner for spatial accelerators with meta reinforcement learning. This repository contains the source code for Poros.

Setup

Download the Poros source code

git clone https://github.com/FuyuWang/Poros.git

Create virtual environment through anaconda

conda create --name PorosEnv python=3.8
conda activate PorosEnv

Install packages

pip install -r requirements.txt

Install MAESTRO

mkdir cost_model
python build.py

Run Poros

Run the Decision procedure from scratch

sh ./run_decison.sh

Run the Decision procedure with the Learning procedure

./run_decison_learning.sh