ezRay 1.1.10

A simple implementation of the ray framework for method agnostic task paralellization.

Get Started

pip install ezRay

Quick Start Guide

from ezRay import MultiCoreExecutionTool
import ray

# configure ezRay
instance_metadata:dict = {
    'num_cpus': 4,              # number of cpus to use
    'num_gpus': 0,              # number of gpus to use
    'address': None,            # remote cluster address. None for local.
    }

# setup ezRay
MultiCore = MultiCoreExecutionTool(instance_metadata = instance_metadata)

# launch ray dashboard (optional)
MultiCore.launch_dashboard()

# define any task
def do_something(foo:int, bar:int) -> int:
    return foo + bar

# or use a ray.remote object
@ray.remote
def do_something_remote(foo:int, bar:int) -> int:
    return foo - bar

# prepare your data in a dictionary. They keys work as identifiers, while the values should be dictionaries matching the function signature.
data = {
    1:{'foo' : 0, 'bar' : 1},
    2:{'foo' : 1, 'bar' : 2},
    3:{'foo' : 2, 'bar' : 3}
    }

# pass the data to ezRay
MultiCore.update_data(data)

# run the task
MultiCore.run(do_something)

# get the results
results_first_task = MultiCore.get_results()

## prepare for the next run
# this will automaticall archive current results
# alternatively you can use MultiCore.archive_results()
MultiCore.next()

# run a second task
MultiCore.run(do_something_remote)

# get current results
results_second_task = MultiCore.get_results()

# get the archived results
archive = MultiCore.get_archive()

Documentation

Pending, sry. No time. However, check out the sandbox in the examples folder or the docstrings in the code.