Module utilities.backend_management

Utility functionality for managing backends

To check a standard backend, use backend_available(). To check a non-standard backend specification, use check_backend().

To build a shim class that raises an error on instantiation, for when a required backend is not available, use missing_backend_shim().

Functions overview

`backend_available`(*backend_names)	Report if a backend is available for use
`check_backend`(backend_name[, ...])	Check if a backend is available, and register it in a list of available backends
`check_samna_available`()	check_samna_available controls if samna package is "installed" and "usable" The default `backend_available()` operation cannot correctly identifies the samna availability.
`check_torch_cuda_available`()
`list_backends`()	Print a list of computational backends available in this session
`missing_backend_error`(class_name, backend_name)	Raise a `ModuleNotFoundError` exception, with information about a missing backend
`missing_backend_shim`(class_name, backend_name)	Make a class constructor that raises an error about a missing backend
`torch_version_satisfied`([req_major, ...])	Check if the installed version of torch satisfies a minimum version requirement

Functions

utilities.backend_management.backend_available(*backend_names) → bool[source]

Report if a backend is available for use

This function returns immediately if the named backend has already been checked previously. Otherwise, if the backend is “standard”, then it will be checked for availability. If the backend is non-standard, it cannot be checked automatically. In that case you must use check_backend() directly.

Parameters:

backend_name0 (str) – A backend to check
backend_name1 (str) – A backend to check
... (str) – A backend to check

Returns:

True iff the backend is available for use

Return type:

bool

utilities.backend_management.check_backend(backend_name: str, required_modules: Tuple[str] | List[str] | None = None, check_flag: bool = True) → bool[source]

Check if a backend is available, and register it in a list of available backends

Parameters:

backend_name (str) – The name of this backend to check for and register
required_modules (Optional[List[str]]) – A list of required modules to search for. If None (default), check the backend name
check_flag (bool) – A manual check that can be performed externally, to see if the backend is available

Returns:

The backend is available

Return type:

bool

utilities.backend_management.check_samna_available() → bool[source]

check_samna_available controls if samna package is “installed” and “usable” The default backend_available() operation cannot correctly identifies the samna availability. In the case that one installed samna and then uninstalled via pip * pip install samna then pip uninstall samna samna leaves a trace and import samna does not raise an error even though the package is not usable.

Returns:: true if samna package available
Return type:: bool

utilities.backend_management.check_torch_cuda_available() → bool[source]

utilities.backend_management.list_backends()[source]: Print a list of computational backends available in this session

utilities.backend_management.missing_backend_error(class_name: str, backend_name: str)[source]

Raise a ModuleNotFoundError exception, with information about a missing backend

Parameters:

class_name (str) – Name of a class which is unavailable
backend_name (str) – “User-facing” of the backend which is unavailable

Raises:

ModuleNotFoundError – Describe the missing backend

utilities.backend_management.missing_backend_shim(class_name: str, backend_name: str)[source]

Make a class constructor that raises an error about a missing backend

Examples

Generate a LIFTorch class shim, that will raise an error on instantiation.

>>> LIFTorch = missing_backend_shim('LIFTorch', 'torch')
>>> LIFTorch((3,), tau_syn = 10e-3)
ModuleNotFoundError: Missing the `torch` backend. `LIFTorch` objects, and others relying on `torch` are not available.

Parameters:

class_name (str) – The intended class name
backend_name (str) – The required backend that is missing

Returns:

A class that raises an error on construction

Return type:

Class

utilities.backend_management.torch_version_satisfied(req_major: int = 0, req_minor: int = 0, req_patch: int = 0) → bool[source]

Check if the installed version of torch satisfies a minimum version requirement

i.e.: torch 2.0.0 >= 1.12.0 : True torch 1.12.0 >= 1.12.0 : True torch 1.11.0 >= 1.12.0 : False

Parameters:

req_major (int) – The minimum major version required
req_minor (int) – The minimum minor version required
req_patch (int) – The minimum patch version required

Returns:

The installed version of torch satisfies the minimum version requirement

Return type:

bool