Generator Functions

As described in the API, the gen_f is called by a libEnsemble worker via the following:

out, persis_info, calc_status = gen_f(H[gen_specs["in"]][sim_ids_from_allocf], persis_info, gen_specs, libE_info)

In practice, most gen_f function definitions written by users resemble:

def my_generator(H, persis_info, gen_specs, libE_info):

Where H is a selection of the History array, determined by sim IDs from the alloc_f, persis_info is a dictionary containing state information, gen_specs is a dictionary containing pre-defined parameters for the gen_f, and libE_info is a dictionary containing libEnsemble-specific entries. Valid generator functions can accept a subset of the above parameters. See the API above for more detailed descriptions of the parameters.

Note

If the gen_f is a persistent generator, then gen_specs["in"] only specifies the fields to send when the gen_f is first called. Use gen_specs["persis_in"] to specify fields to send back to the generator throughout runtime.

Typically users start by extracting their custom parameters initially defined within gen_specs["user"] in the calling script and defining a local History array based on the datatype in gen_specs["out"], to be returned. For example:

batch_size = gen_specs["user"]["batch_size"]
local_H_out = np.zeros(batch_size, dtype=gen_specs["out"])

This array should be populated by whatever values are generated within the function. Finally, this array should be returned to libEnsemble alongside persis_info if it was passed in:

return local_H_out, persis_info

Between the output array definition and the function returning, any level and complexity of computation can be performed. Users are encouraged to use the executor to submit applications to parallel resources if necessary, or plug in components from any other libraries to serve their needs.

Note

State gen_f information like checkpointing should be appended to persis_info.

Persistent Generators

While non-persistent generators return after completing their calculation, persistent generators receive work units, perform computations, and communicate results directly to the manager in a loop. A persistent generator returns either when explicitly instructed by the manager, or by exiting its main loop based on some condition. The allocation function can determine what to do once a persistent generator finishes, such as ending the ensemble.

The calling worker becomes a dedicated persistent worker. A gen_f is initiated as persistent by the alloc_f.

Many users prefer persistent generators since they do not need to be re-initialized every time their past work is completed and evaluated by a simulation, and can evaluate returned simulation results over the course of an entire libEnsemble routine as a single function instance. The APOSMM optimization generator function included with libEnsemble is persistent so it can maintain multiple local optimization subprocesses based on results from complete simulations.

Functions for a persistent generator to communicate directly with the manager are available in the libensemble.tools.persistent_support class. Additional necessary resources are the status tags STOP_TAG, PERSIS_STOP, EVAL_GEN_TAG, and FINISHED_PERSISTENT_GEN_TAG from libensemble.message_numbers. Return values from the persistent_support functions are compared to these tags to determine when the generator should break its loop and return.

A PersistentSupport class instance should resemble:

my_support = PersistentSupport(libE_info, EVAL_GEN_TAG)

Implementing functions from the above class is relatively simple:

libensemble.tools.persistent_support.PersistentSupport.send(self, output, calc_status=0, keep_state=False)

Send message from worker to manager.

Parameters:

• output (ndarray[Any, dtype[ScalarType]]) – Output array to be sent to manager

• calc_status (int) – Optional, Provides a task status

• keep_state – Optional, If True the manager will not modify its record of the workers state (usually the manager changes the worker’s state to inactive, indicating the worker is ready to receive more work, unless using active receive mode).

Return type:

None

This function call typically resembles:

my_support.send(local_H_out[selected_IDs])

Note that this function has no return.

libensemble.tools.persistent_support.PersistentSupport.recv(self, blocking=True)

Receive message to worker from manager.

Parameters:

blocking (bool) – Optional, If True (default), will block until a message is received.

Returns:

message tag, Work dictionary, calc_in array

Return type:

(<class ‘int’>, <class ‘dict’>, numpy.ndarray[Any, numpy.dtype[+ScalarType]])

This function call typically resembles:

tag, Work, calc_in = my_support.recv()

if tag in [STOP_TAG, PERSIS_STOP]:
    cleanup()
    break

The logic following the function call is typically used to break the persistent generator’s main loop and return.

libensemble.tools.persistent_support.PersistentSupport.send_recv(self, output, calc_status=0)

Send message from worker to manager and receive response.

Parameters:

• output (ndarray[Any, dtype[ScalarType]]) – Output array to be sent to manager

• calc_status (int) – Optional, Provides a task status

Returns:

message tag, Work dictionary, calc_in array

Return type:

(<class ‘int’>, <class ‘dict’>, numpy.ndarray[Any, numpy.dtype[+ScalarType]])

This function performs both of the previous functions in a single statement. Its usage typically resembles:

tag, Work, calc_in = my_support.send_recv(local_H_out[selected_IDs])
if tag in [STOP_TAG, PERSIS_STOP]:
    cleanup()
    break

Once the persistent generator’s loop has been broken because of the tag from the manager, it should return with an additional tag:

return local_H_out, persis_info, FINISHED_PERSISTENT_GEN_TAG

See calc_status for more information about the message tags.

Active receive mode

By default, a persistent worker (generator in this case) models the manager/worker communications of a regular worker (i.e., the generator is expected to alternately receive and send data in a ping pong fashion). To have an irregular communication pattern, a worker can be initiated in active receive mode by the allocation function (see start_only_persistent). In this mode, the persistent worker will always be considered ready to receive more data (e.g.,~ evaluation results). It can also send to the manager at any time.

The user is responsible for ensuring there are no communication deadlocks in this mode. Note that in manager/worker message exchanges, only the worker-side receive is blocking by default (a non-blocking option is available).

Cancelling Simulations

Previously submitted simulations can be cancelled by sending a message to the manager.

To do this a PersistentSupport helper function is provided.

libensemble.tools.persistent_support.PersistentSupport.request_cancel_sim_ids(self, sim_ids)

Request cancellation of sim_ids

Parameters:

sim_ids (List[int]) – A list of sim_ids to cancel

A message is sent to the manager to mark requested sim_ids as cancel_requested

If a generated point is cancelled by the generator before it has been given to a worker for evaluation, then it will never be given. If it has already returned from the simulation, then results can be returned, but the cancel_requested field remains as True. However, if the simulation is running when the manager receives the cancellation request, a kill signal will be sent to the worker. This can be caught and acted upon by a user function, otherwise it will be ignored.

The Borehole Calibration tutorial gives an example of the capability to cancel pending simulations.

Modification of existing points

To change existing fields of the history array, the generator can initialize an output array where the dtype contains the sim_id and the fields to be modified (in place of gen_specs["out"]), and then send this output array to the manager (as with regular communications). Any such array received by the manager will overwrite the specific fields for the given sim_ids. If the changes do not correspond with newly generated points, then the generator needs to communicate to the manager that it is not ready to receive completed evaluations. Send to the manager with the keep_state argument set to True.

For example, the cancellation function request_cancel_sim_ids could be replicated by the following (where sim_ids_to_cancel is a list of integers):

# Send only these fields to existing H rows and libEnsemble will slot in the change.
H_o = np.zeros(len(sim_ids_to_cancel), dtype=[("sim_id", int), ("cancel_requested", bool)])
H_o["sim_id"] = sim_ids_to_cancel
H_o["cancel_requested"] = True
ps.send(H_o, keep_state=True)

Generator initiated shutdown

If using a supporting allocation function, the generator can prompt the ensemble to shutdown by simply exiting the function (e.g., on a test for a converged value). For example, the allocation function start_only_persistent closes down the ensemble as soon a persistent generator returns. The usual return values should be given.

Examples

Examples of non-persistent and persistent generator functions can be found here.