badcrossbar package#

Subpackages#

Submodules#

badcrossbar.check module#

badcrossbar.check.crossbar_requirements(resistances, applied_voltages, r_i_word_line, r_i_bit_line, **kwargs)[source]#

Checks if crossbar variables satisfy all requirements.

Parameters

  • resistances (Union[_SupportsArray[dtype], _NestedSequence[_SupportsArray[dtype]], bool, int, float, complex, str, bytes, _NestedSequence[Union[bool, int, float, complex, str, bytes]]]) – Resistances of crossbar devices.

  • applied_voltages (Union[_SupportsArray[dtype], _NestedSequence[_SupportsArray[dtype]], bool, int, float, complex, str, bytes, _NestedSequence[Union[bool, int, float, complex, str, bytes]]]) – Applied voltages.

  • r_i_word_line – Interconnect resistance of the word line segments.

  • r_i_bit_line – Interconnect resistance of the bit line segments.

Return type

tuple[ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]], ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]]

Returns

Potentially modified resistances and applied voltages.

badcrossbar.check.match_shape(**kwargs)[source]#

Checks if arrays have matching dimensions.

Parameters

**kwargs – Arrays and the dimension along which they should be matched.

Raises

ValueError – If any of the arrays do not match specified dimensions.

badcrossbar.check.n_dimensional(array, n_list=[2], name='array')[source]#

Checks that array is n-dimensional.

Parameters

  • array (ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]) – Array.

  • n_list (list[int]) – Possible number of dimensions.

  • name (str) – Name of the variable.

Raises

TypeError – If array is not n-dimensional.

badcrossbar.check.non_empty(array, name='array')[source]#

Checks that array is not empty.

Parameters

  • array (ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]) – Array.

  • name (str) – Name of the array.

Raises

ValueError – If the array is empty.

badcrossbar.check.non_infinite_array(array, name='array')[source]#

Checks if all the elements of the array are non-infinite.

Parameters

  • array (ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]) – Array.

  • name (str) – Name of the array.

Raises

ValueError – If the array contains positive or negative infinities.

badcrossbar.check.non_negative_array(array, name='array')[source]#

Checks if all the elements of the array are non-negative.

Parameters

  • array (ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]) – Array.

  • name (str) – Name of the array.

Raises

ValueError – If the array contains negative values.

badcrossbar.check.non_negative_number(value, name='number')[source]#

Checks if the number is negative.

Parameters

  • value (float) – Number.

  • name (str) – Name of the number.

Raises

ValueError – If the number is negative.

badcrossbar.check.not_none(**kwargs)[source]#

Confirms that at least one of the items is not None.

Parameters

**kwargs – Items of arbitrary type.

Return type

dict[str, Any]

Returns

Items that are not None.

Raises

ValueError – If all of the items are None.

badcrossbar.check.number(value, name='variable')[source]#

Checks if the variable is a number.

Parameters

  • value (Any) – Variable of arbitrary type.

  • name (str) – Name of the variable.

Raises

TypeError – If the variable is not int or float.

badcrossbar.check.numeric_array(array, name='array')[source]#

Checks that array only contains numbers.

Parameters

  • array (ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]) – Array.

  • name (str) – Name of the array.

Raises

TypeError – If array contains non-number elements.

badcrossbar.check.plotting_requirements(device_branch_vals=None, word_line_branch_vals=None, bit_line_branch_vals=None, word_line_node_vals=None, bit_line_node_vals=None, branches=True)[source]#

Checks if arrays containing branch or node values satisfy all requirements.

Parameters

  • device_branch_vals (Optional[ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]]) – Values associated with crossbar devices.

  • word_line_branch_vals (Optional[ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]]) – Values associated with the interconnect segments along the word lines.

  • bit_line_branch_vals (Optional[ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]]) – Values associated with the interconnect segments along the bit lines.

  • word_line_node_vals (Optional[ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]]) – Values associated with the nodes on the word lines.

  • bit_line_node_vals (Optional[ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]]) – Values associated with the nodes on the bit lines.

  • branches (bool) – If True, it is assumed that branch values are passed. Otherwise, node values are expected.

Return type

ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]

Returns

Potentially modified branch or nodes values.

badcrossbar.check.short_circuit(resistances, r_i_word_line, r_i_bit_line)[source]#

Checks if crossbar will be short-circuited.

This refers to a theoretical scenario when there exists a path of zero resistance in a crossbar.

Parameters

  • resistances (ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]) – Resistances of crossbar devices.

  • r_i_word_line (float) – Interconnect resistance of the word line segments.

  • r_i_bit_line (float) – Interconnect resistance of the bit line segments.

Raises

ValueError – If any of the devices have zero resistance.

badcrossbar.compute module#

badcrossbar.compute.compute(applied_voltages, resistances, r_i=None, r_i_word_line=None, r_i_bit_line=None, **kwargs)[source]#

Computes branch currents and node voltages of a crossbar.

Parameters

  • applied_voltages (Union[_SupportsArray[dtype], _NestedSequence[_SupportsArray[dtype]], bool, int, float, complex, str, bytes, _NestedSequence[Union[bool, int, float, complex, str, bytes]]]) – Applied voltages. Voltages must be supplied in an array of shape m x p, where m is the number of word lines and p is the number of examples (sets of voltages applied one by one).

  • resistances (Union[_SupportsArray[dtype], _NestedSequence[_SupportsArray[dtype]], bool, int, float, complex, str, bytes, _NestedSequence[Union[bool, int, float, complex, str, bytes]]]) – Resistances of crossbar devices. Resistances must be supplied in an array of shape m x n, where n is the number of bit lines.

  • r_i (Optional[float]) – Interconnect resistance of the word and bit line segments. If None, r_i_word_line and r_i_bit_line are used instead.

  • r_i_word_line (Optional[float]) – Interconnect resistance of the word line segments.

  • r_i_bit_line (Optional[float]) – Interconnect resistance of the bit line segments.

  • **node_voltages – If False, None is returned instead of node voltages.

  • **all_currents – If False, only output currents are returned, while all the other ones are set to None.

Return type

Solution

Returns

Branch currents and node voltages of the crossbar. Field currents is a named tuple itself with fields output, device, word_line and bit_line and contains output currents, as well as currents flowing through the devices and interconnect segments of the word and bit lines. Field voltages is a named tuple itself with fields word_line and bit_line and contains the voltages at the nodes on the word and bit lines. currents.output is an array of shape p x n, while all the others are arrays of shape m x n if p == 1, or arrays of shape m x n x p if p > 1.

badcrossbar.plot module#

badcrossbar.plot.branches(device_vals=None, word_line_vals=None, bit_line_vals=None, currents=None, **kwargs)[source]#

Plots a crossbar array and colors its branches according to the values passed. The diagram is saved as a PDF file.

If currents is passed, then it is used to plot the currents in the branches. Otherwise, at least one of {device_vals, word_line_vals, bit_line_vals} has to be passed.

Parameters

  • device_vals (Union[_SupportsArray[dtype], _NestedSequence[_SupportsArray[dtype]], bool, int, float, complex, str, bytes, _NestedSequence[Union[bool, int, float, complex, str, bytes]], None]) – Values associated with crossbar devices.

  • word_line_vals (Union[_SupportsArray[dtype], _NestedSequence[_SupportsArray[dtype]], bool, int, float, complex, str, bytes, _NestedSequence[Union[bool, int, float, complex, str, bytes]], None]) – Values associated with the interconnect segments along the word lines.

  • bit_line_vals (Union[_SupportsArray[dtype], _NestedSequence[_SupportsArray[dtype]], bool, int, float, complex, str, bytes, _NestedSequence[Union[bool, int, float, complex, str, bytes]], None]) – Values associated with the interconnect segments along the bit lines.

  • currents (Optional[Currents]) – Crossbar branch currents. It should have fields device, word_line and bit_line that contain currents flowing through the devices and interconnect segments of the word and bit lines (at least one of them should be not None).

  • **default_color – Normalized RGB values of the nodes and certain types of branches if their values are not provided.

  • **wire_scaling_factor – Scaling factor for the width of the word and bit lines.

  • **device_scaling_factor – Scaling factor for the width of the devices. Also scales the nodes.

  • **node_scaling_factor – Scaling factor for the diameter of the nodes which is combined with device_scaling_factor. For example, if one wanted to only scale the device width by a factor of 2, but keep the node diameter the same, arguments device_scaling_factor = 2 and node_scaling_factor = 1/2 would have to be passed.

  • **axis_label – Axis label of the color bar.

  • **low_rgb – Normalized RGB value associated with the lower limit.

  • **zero_rgb – Normalized RGB value associated with the value of zero.

  • **high_rgb – Normalized RGB value associated with the upper limit.

  • **allow_overwrite – If True, can overwrite existing PDF files with the same name.

  • **filename – Filename, excluding PDF extension.

  • **device_type – Device type to be drawn. One of {‘memristor’, ‘memristor_2’, ‘resistor_usa’, ‘resistor_europe’}.

  • **significant_figures – Number of significant figures to use for the limits of the color bar.

  • **round_crossings – Because the circuit of a crossbar array is non-planar, the 2D diagram of it will have some wire crossings. If round_crossings is False, these crossings will be drawn as straight lines. Otherwise, they will be drawn as semicircles.

  • **width – Width of the diagram in millimeters.

badcrossbar.plot.nodes(word_line_vals=None, bit_line_vals=None, voltages=None, **kwargs)[source]#

Plots a crossbar array and colors its nodes according to the values passed. The diagram is saved as a PDF file.

If voltages is passed, then it is used to plot the voltages on the nodes. Otherwise, at least one of {word_line_vals, bit_line_vals} has to be passed.

Parameters

  • word_line_vals (Union[_SupportsArray[dtype], _NestedSequence[_SupportsArray[dtype]], bool, int, float, complex, str, bytes, _NestedSequence[Union[bool, int, float, complex, str, bytes]], None]) – Values associated with the nodes on the word lines.

  • bit_line_vals (Union[_SupportsArray[dtype], _NestedSequence[_SupportsArray[dtype]], bool, int, float, complex, str, bytes, _NestedSequence[Union[bool, int, float, complex, str, bytes]], None]) – Values associated with the nodes on the bit lines.

  • voltages (Optional[Voltages]) – Crossbar node voltages. It should have fields word_line and bit_line that contain the potentials at the nodes on the word and bit lines (at least one of them should be not None).

  • **default_color – Normalized RGB values of the branches and certain type of nodes if its values are not provided.

  • **wire_scaling_factor – Scaling factor for the width of the word and bit lines.

  • **device_scaling_factor – Scaling factor for the width of the devices.

  • nodes. (Also scales the) –

  • **node_scaling_factor – Scaling factor for the diameter of the nodes which is combined with device_scaling_factor. For example, if one wanted to only scale the device width by a factor of 2, but keep the node diameter the same, arguments device_scaling_factor = 2 and node_scaling_factor = 1/2 would have to be passed.

  • **axis_label – Axis label of the color bar.

  • **low_rgb – Normalized RGB value associated with the lower limit.

  • **zero_rgb – Normalized RGB value associated with the value of zero.

  • **high_rgb – Normalized RGB value associated with the upper limit.

  • **allow_overwrite – If True, can overwrite existing PDF files with the same name.

  • **filename – Filename, excluding PDF extension.

  • **device_type – Device type to be drawn. One of {‘memristor’, ‘memristor_2’, ‘resistor_usa’, ‘resistor_europe’}.

  • **significant_figures – Number of significant figures to use for the limits of the color bar.

  • **round_crossings – Because the circuit of a crossbar array is non-planar, the 2D diagram of it will have some wire crossings. If round_crossings is False, these crossings will be drawn as straight lines. Otherwise, they will be drawn as semicircles.

  • **width – Width of the diagram in millimeters.

badcrossbar.utils module#

badcrossbar.utils.arrays_shape(*arrays)[source]#

Returns the shape of the first array that is not None.

Parameters

arrays (ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]) – Arrays.

Returns

Shape.

badcrossbar.utils.average_if_3D(array)[source]#

If array is 3D, it is averaged along the third axis.

Parameters

array (ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]) – 2D or 3D array.

Return type

ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]

Returns

2D array.

badcrossbar.utils.distributed_array(flattened_array, model_array)[source]#

Reshapes flattened array.

Parameters

  • flattened_array (ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]) – An array whose each column contains a flattened array.

  • model_array (ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]) – An array whose shape is used for reshaping.

Return type

ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]

Returns

Array or a list of arrays in specified shape.

badcrossbar.utils.load_pickle(path, sanitize=True)[source]#

Loads pickle file.

Parameters

  • path (str) – Path to the pickle file, including extension.

  • sanitize (bool) – If True, sanitizes the filename by removing illegal characters and making the path compatible with the operating system.

Returns

Extracted contents.

badcrossbar.utils.save_pickle(variable, path, allow_overwrite=False, sanitize=True)[source]#

Saves variable to a pickle file.

Parameters

  • variable – Variable to be saved.

  • path (str) – Path to the pickle file, excluding extension.

  • allow_overwrite (bool) – If False, will not check for existing files with the same name and will overwrite if such files exist.

  • sanitize (bool) – If True, sanitizes the filename by removing illegal characters and making the path compatible with the operating system.

badcrossbar.utils.squeeze_third_axis(array)[source]#

Removes third axis of ndarray if it has shape of 1.

Parameters

array (ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]) – 3D array.

Return type

ndarray[Any, dtype[TypeVar(ScalarType, bound= generic, covariant=True)]]

Returns

2D or 3D array.

badcrossbar.utils.unique_path(path, extension='pdf', sanitize=True)[source]#

Append a number to the path, if it is not unique.

Parameters

  • path (str) – Path of the filename without the extension.

  • extension (str) – File extension.

  • sanitize (bool) – If True, sanitizes the filename by removing illegal characters and making the path compatible with the operating system.

Return type

str

Returns

Unique path.

Module contents#