A pipeline system for efficient execution.
Project description
Pyturbo Package
Author: Lijun Yu
Email: lijun@lj-y.com
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.
Get Started
from pyturbo import ReorderStage, Stage, System
class Stage1(Stage): # Define a stage
def __init__(self, resources):
... # Optional: set resources and 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
system.add_job(...) # Submit one job
finished_job = system.result_queue.get() # Wait for result
system.end() # End system
Demo
See demo.py for an example implementation.
Development Modes
See develop.md
Version History
See version.md.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
py-turbo-0.2.3.tar.gz
(8.1 kB
view hashes)
Built Distribution
py_turbo-0.2.3-py3-none-any.whl
(22.6 kB
view hashes)