Configuration

Sionna PHY’s configuration API. It can be used to set global variables which can be used by all of its modules and functions.

class sionna.phy.config.Config

Bases: object

Sionna PHY Configuration Class

This singleton class is used to define global configuration variables and random number generators that can be accessed from all modules and functions. It is instantiated immediately and its properties can be accessed as sionna.phy.config.desired_property.

Methods

torch_rng(device: str | None = None) → torch._C.Generator

torch.Generator : PyTorch random number generator for the specified device

Parameters:

device (str | None) – Device name (e.g., 'cpu', 'cuda:0'). If None, device is used.

Example

from sionna.phy import config
config.seed = 42 # Set seed for deterministic results

# Use generator instead of torch.randn
noise = torch.randn([4], generator=config.torch_rng())

Attributes

property available_devices: list[str]

list of str : List of available compute devices

property cdtype: torch.dtype

torch.dtype : Default PyTorch dtype for complex floating point numbers

property device: str

str : Get/set the device for computation (e.g., 'cpu', 'cuda:0')

property dtype: torch.dtype

torch.dtype : Default PyTorch dtype for real floating point numbers

property np_cdtype: numpy.dtype

np.dtype : Default NumPy dtype for complex floating point numbers

property np_dtype: numpy.dtype

np.dtype : Default NumPy dtype for real floating point numbers

property np_rng: numpy.random._generator.Generator

np.random.Generator : NumPy random number generator

Example

from sionna.phy import config
config.seed = 42 # Set seed for deterministic results

# Use generator instead of np.random
noise = config.np_rng.normal(size=[4])

property precision: Literal['single', 'double']

"single" (default) | "double" : Default precision used for all computations

The "single" option represents real-valued floating-point numbers using 32 bits, whereas the "double" option uses 64 bits. For complex-valued data types, each component of the complex number (real and imaginary parts) uses either 32 bits (for "single") or 64 bits (for "double").

property py_rng: random.Random

random.Random : Python random number generator

Example

from sionna.phy import config
config.seed = 42 # Set seed for deterministic results

# Use generator instead of random
val = config.py_rng.randint(0, 10)

property seed: int | None

None (default) | int : Get/set seed for all random number generators

All random number generators used internally by Sionna can be configured with a common seed to ensure reproducibility of results. It defaults to None which implies that a random seed will be used and results are non-deterministic.

Example

# This code will lead to deterministic results
from sionna.phy import config
from sionna.phy.mapping import BinarySource
config.seed = 42
print(BinarySource()([10]))