Cache

from numpy import array

from gemseo import create_discipline
from gemseo.caches.hdf5_cache import HDF5Cache

Create a discipline

discipline = create_discipline("AnalyticDiscipline", expressions={"z": "x+y"})

Set the cache policy to store all executions:

discipline.set_cache("HDF5Cache", hdf_file_path="file.h5")  # on the disk
discipline.set_cache("MemoryFullCache")  # in memory

Set the simple cache policy to store the last execution in memory:

discipline.set_cache("SimpleCache")  # default option

Export cache to dataset:

input_data = {"x": array([1.0]), "y": array([2.0])}
discipline.execute(input_data)
dataset = discipline.cache.to_dataset()

Cache inputs and outputs in an HDF5 file:

input_data = {"x": array([1.0]), "y": array([2.0])}
output_data = {"z": array([3.0])}
cache = HDF5Cache(hdf_file_path="file.h5", hdf_node_path="node")
cache.cache_outputs(input_data, output_data)

Get cached data:

last_entry = cache.last_entry
last_cached_input_data = last_entry.inputs
last_cached_output_data = last_entry.outputs
n_entries = len(cache)

Get outputs and jacobian if data are cached, else None:

_, output_data, jac_data = cache[input_data]