Source code for spynnaker.pyNN.models.current_sources.abstract_current_source

# Copyright (c) 2017 The University of Manchester
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import annotations
from enum import Enum
from typing import Mapping, Optional, Sequence, Union, TYPE_CHECKING
from typing_extensions import TypeAlias
from spinn_utilities.abstract_base import AbstractBase, abstractmethod
from spinn_front_end_common.interface.ds import DataType
if TYPE_CHECKING:
    from spynnaker.pyNN.models.populations import Population, PopulationBase
    from spynnaker.pyNN.models.neuron.abstract_population_vertex import (
        AbstractPopulationVertex)

#: General type of parameters to current sources.
#: Individual parameters will only be one of these!
CurrentParameter: TypeAlias = Union[int, float, Sequence[int], Sequence[float]]


class CurrentSourceIDs(Enum):
    """
    Hashes of the current sources currently supported
    """
    NO_SOURCE = 0
    DC_SOURCE = 1
    AC_SOURCE = 2
    STEP_CURRENT_SOURCE = 3
    NOISY_CURRENT_SOURCE = 4
    N_SOURCES = 4


class AbstractCurrentSource(object, metaclass=AbstractBase):
    """
    A simplified version of the PyNN class, since in most cases we work
    out the actual offset value on the SpiNNaker machine itself based on
    the parameters during the run.
    """
    __slots__ = (
        "__app_vertex",
        "__population")

    def __init__(self) -> None:
        self.__app_vertex: Optional[AbstractPopulationVertex] = None
        self.__population: Optional[Population] = None

[docs] def inject_into(self, cells: PopulationBase): """ Inject this source into the specified population cells. :param PopulationBase cells: The cells to inject the source into """ # Call the population method to pass the source in cells.inject(self)
[docs] def set_app_vertex(self, vertex: AbstractPopulationVertex): """ Set the application vertex associated with the current source. :param AbstractPopulationVertex vertex: The population vertex """ self.__app_vertex = vertex
@property def app_vertex(self) -> Optional[AbstractPopulationVertex]: """ The application vertex associated with the current source. :rtype: AbstractPopulationVertex """ return self.__app_vertex
[docs] def set_population(self, population: Population): """ Set the population associated with the current source. :param ~spynnaker.pyNN.models.populations.Population population: """ self.__population = population
@property def population(self) -> Optional[Population]: """ The population associated with the current source. :rtype: ~spynnaker.pyNN.models.populations.Population """ return self.__population
[docs] @abstractmethod def set_parameters(self, **parameters: CurrentParameter): """ Set the current source parameters. :param parameters: the parameters to set """ raise NotImplementedError
@property @abstractmethod def parameters(self) -> Mapping[str, CurrentParameter]: """ The parameters of the current source. :rtype: dict(str, Any) """ raise NotImplementedError @property @abstractmethod def parameter_types(self) -> Mapping[str, DataType]: """ The parameter types for the current source. :rtype: dict(str, ~.DataType) """ raise NotImplementedError @property @abstractmethod def current_source_id(self) -> int: """ The ID of the current source. :rtype: int """ raise NotImplementedError
[docs] @abstractmethod def get_sdram_usage_in_bytes(self) -> int: """ The SDRAM usage in bytes of the current source. :rtype: int """ raise NotImplementedError