Choosing an Executor for a Task (Electron)#

Within a lattice, every electron can use a different executor. These exectors can have different capabilities: different hardware, different computation strategies, and so on.

Executors are plugins. Any executor plugins found by the dispatcher are imported as classes in the covalent.executor name space.

Prerequisites#

  1. Start the Covalent server.

  2. Import covalent.

[1]:

import covalent as ct

Procedure#

Initialize an executor with its class:

[2]:

executor1 = ct.executor.LocalExecutor()

  1. Assign the executor to an electron using the keyword argument executor:

[3]:

@ct.electron(executor=executor1)
def identity(x):
    return x

Another way to specify an executor is by the short name of the executor plugin.

In this case, the short name of the LocalExecutor is local:

[4]:

@ct.electron(executor="local")
def square(x):
    return x * x

  1. Construct rest of the workflow as usual within a lattice:

[5]:

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

  1. Submit the workflow using the dispatch method:

[6]:

dispatch_id = ct.dispatch(workflow)(a=2)
print(dispatch_id)

5f0e220f-35dc-4a75-92c6-a00176f530e2

Dispatching the workflow generates a dispatch ID.

Use this ID to query the status of the task and retrieve the results:

[7]:

output = ct.get_result(dispatch_id=dispatch_id, wait=True)
print(output)


Lattice Result
==============
status: COMPLETED
result: 4
input args: []
input kwargs: {'a': '2'}
error: None

start_time: 2023-01-31 21:20:27.871124
end_time: 2023-01-31 21:20:28.054424

results_dir: /Users/mini-me/agnostiq/covalent/doc/source/how_to/execution/results
dispatch_id: 5f0e220f-35dc-4a75-92c6-a00176f530e2

Node Outputs
------------
identity(0): 2
:parameter:2(1): 2
square(2): 4

See Also#

Getting the Results of an Electron Execution

Getting Results of Previous Runs