Skip to content

CMU-INF-DIVA/pyturbo

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

72 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Pyturbo Package

PyPI version Downloads Publish to PyPI

Author: Lijun Yu

Email: [email protected]

A pipeline system for efficient execution.

Installation

pip install py-turbo

Introduction

Pyturbo utilizes multiple level of abstract to efficiently execute parallel tasks.

  • Worker: a process.
  • Stage: a group of peer workers processing the same type of tasks.
  • Task: a data unit transferred between stages. At each stage, a task is processed by one worker and will result in one or multiple downstream tasks.
  • Pipeline: a set of sequential stages.
  • Job: a data unit for a pipeline, typically a wrapped task for the first stage.
  • Result: output of a job processed by one pipeline, typically a set of output tasks from the last stage.
  • System: a set of peer pipelines processing the same type of jobs.

abstract.png

Get Started

from pyturbo import ReorderStage, Stage, System

class Stage1(Stage): # Define a stage

    def allocate_resource(self, resources, ...):
        ... # Optional: split resources and determine number of workers.

    def process(self, task):
        ... # Process function for each worker process. Returns one or a series of downstream tasks.

... # Repeat for Stage2, Stage3

class Stage4(ReorderStage): # Define a reorder stage, typically for the final stage

    def get_sequence_id(self, task):
        ... # Return the order of each task. Start from 0.

    def process(self, task):
        ...

class MySystem(System):

    def get_stages(self, resources):
        ... # Define the stages in a pipeline with given resources.

    def get_results(self, results_gen):
        ... # Define how to extract final results from output tasks.

def main():
    system = MySystem(num_pipeline) # Set debug=True to run in a single process
    system.start() # Build and start system
    jobs = [...]
    system.add_jobs(jobs) # Submit jobs
    for job in system.wait_jobs(len(jobs)):
        print(job.results) # Process result
    system.end() # End system

Options

See options.md

Demo

abstract.png

See demo.py for an example implementation.

Version History

See version.md.