simpeg.meta.DaskMetaSimulation

class simpeg.meta.DaskMetaSimulation(simulations, mappings, client)

Bases: MetaSimulation

Dask Distributed version of simulation of simulations.

This class makes use of dask.distributed module to provide concurrency, executing the internal simulations in parallel. This class is meant to be a (mostly) drop in replacement for MetaSimulation. If you want to test your implementation, we recommend starting with a small problem using MetaSimulation, then switching it to this class. the serial version of this class is good for testing correctness.

Parameters:

simulations(n_sim) list of simpeg.simulation.BaseSimulation or list of dask.distributed.Future

The list of unique simulations (or futures that would return a simulation) that each handle a piece of the problem.

mappings(n_sim) list of simpeg.maps.IdentityMap or list of dask.distributed.Future

The map for every simulation (or futures that would return a map). Every map should accept the same length model, and output a model appropriate for its paired simulation.

clientdask.distributed.Client, optional

The dask client to use for communication.

Attributes

clean_on_model_update	A list of solver objects to clean when the model is updated
client	The distributed client that handles the internal tasks.
counter	SimPEG Counter object to store iterations and run-times.
deleteTheseOnModelUpdate	HasModel.deleteTheseOnModelUpdate has been deprecated.
mappings	The future mappings paired to each simulation.
model	The inversion model.
needs_model	True if a model is necessary
sensitivity_path	Path to directory where sensitivity file is stored.
simulations	The future list of simulations.
survey	The survey for the simulation.
verbose	Verbose progress printout.

Methods

Jtvec(m, v[, f])	Compute the Jacobian transpose times a vector for the model provided.
Jtvec_approx(m, v[, f])	Approximation of the Jacobian transpose times a vector for the model provided.
Jvec(m, v[, f])	Compute the Jacobian times a vector for the model provided.
Jvec_approx(m, v[, f])	Approximation of the Jacobian times a vector for the model provided.
dpred([m, f])	Predicted data for the model provided.
fields(m)	Create fields for every simulation.
getJtJdiag(m[, W, f])	Return the squared sum of columns of the Jacobian.
make_synthetic_data(m[, relative_error, ...])	Make synthetic data for the model and Gaussian noise provided.
residual(m, dobs[, f])	The data residual.