HDF5 cache

In this example, we will see how to use HDF5Cache.

from __future__ import annotations

from numpy import array

from gemseo import configure_logger
from gemseo.caches.hdf5_cache import HDF5Cache

configure_logger()

<RootLogger root (INFO)>

Import

Let's first import the array and the HDF5Cache classes.

Create

An instance of HDF5Cache can be instantiated with the following statement. The user has to provide the file path of the HDF5 file, as well as the node name, which usually is a discipline name.

Warning

The HDF5Cache relies on some multiprocessing features. When working on Windows, the execution of scripts containing instances of HDF5Cache must be protected by an if __name__ == '__main__': statement. Currently, the use of an HDF5Cache is not supported in parallel on Windows platforms. This is due to the way subprocesses are forked in this architecture. The method DOEScenario.set_optimization_history_backup() is recommended as an alternative.

cache = HDF5Cache(hdf_file_path="my_cache.hdf5", hdf_node_path="node1")

It is possible to see the principal attributes of the cache by printing it, either using a print statement or using the logger:

cache

Name: node1

• Type: HDF5Cache
• Tolerance: 0.0
• Input names: []
• Output names: []
• Length: 0
• HDF file path: my_cache.hdf5
• HDF node path: node1

Cache

In this example, we manually add data in the cache from the data dictionary to illustrate its use. Yet, it has to be noted that a cache can be attached to an Discipline instance, and the user does not have to feed the cache manually. Here, we provide to the cache the data dictionary, and we set x as input and y as output.

cache[{"x": array([1.0])}] = ({"y": array([2.0])}, None)
cache[{"x": array([2.0])}] = ({"y": array([3.0])}, None)
cache

Get all data

We can now print some information from the cache, such as its length:

len(cache)

2

We can also display all the cached data so far.

list(cache)

[CacheEntry(inputs={'x': array([1.])}, outputs={'y': array([2.])}, jacobian={}), CacheEntry(inputs={'x': array([2.])}, outputs={'y': array([3.])}, jacobian={})]

Get last cached data

It is also possible to display the last entry cached, for the inputs and the outputs.

last_entry = cache.last_entry
last_entry.inputs, last_entry.outputs

({'x': array([2.])}, {'y': array([3.])})

Clear the cache

It is also possible to clear the cache, which removes all the data which has been stored so far in the HDF5 file.

cache.clear()
cache

Name: node1

• Type: HDF5Cache
• Tolerance: 0.0
• Input names: []
• Output names: []
• Length: 0
• HDF file path: my_cache.hdf5
• HDF node path: node1