Executing a Workflow (Lattice) Multiple Times#

Execute a lattice multiple times with different parameters by starting the lattice from a loop. Covalent parallelizes the execution of the runs.

Prerequisites#

Start the Covalent server.

Procedure#

  1. Write the lattice. This example takes a single scalar parameter, but the principle is the same for more complex inputs.

[1]:

import covalent as ct

@ct.electron
def identity(x):
    return x

@ct.electron
def square(x):
    return x * x

@ct.lattice
def workflow(a):
    val_1 = identity(x=a)
    return square(x=val_1)

  1. Dispatch the tasks separately:

[2]:

params = [1, 2, 3, 4]
dispatch_ids = [ct.dispatch(workflow)(a=param) for param in params]

dispatch_ids

[2]:

['08a5f022-0bb7-4f70-b7bd-6a6806931e40',
 'd4a3e660-63a2-47b0-8d3e-722b697bcd7d',
 'e34ed7ef-be70-4700-954b-9ed7330b2a86',
 '1cffea94-5d82-4186-a352-0a510610e4cc']

  1. The dispatch IDs are your keys to the results of each run; you will use them to process or view the results. One option is to simply save the dispatch IDs so that you can retrieve the results later:

[3]:

with open("dispatch_ids.txt", 'w') as id_file:
    id_file.writelines("%s\n" % id for id in dispatch_ids)

See Also#

Looping

Executing a Lattice as an Electron (Sublattice)