hgq.layers.snn package

Submodules

hgq.layers.snn.base module

class hgq.layers.snn.base.ATan(alpha: float = 2.0)

Bases: object

Arc-tangent surrogate gradient for a Heaviside spike function.

get_config()

class hgq.layers.snn.base.LIFCell(*args, **kwargs)

Bases: SpikingNeuralCell

Leaky integrate-and-fire RNN cell.

build(input_shape)

call(inputs, states, training=None)

get_config()

Returns the config of the object.

An object config is a Python dictionary (serializable) containing the information needed to re-instantiate it.

class hgq.layers.snn.base.QLIF(*args, **kwargs)

Bases: _QSNN

Quantization-aware leaky integrate-and-fire sequence layer.

get_config()

Returns the config of the object.

An object config is a Python dictionary (serializable) containing the information needed to re-instantiate it.

class hgq.layers.snn.base.QLIFCell(*args, **kwargs)

Bases: QSimpleSNNCell

Quantization-aware leaky integrate-and-fire RNN cell.

property beta_q

build(input_shape)

call(inputs, states, training=None)

get_config()

Returns the config of the object.

An object config is a Python dictionary (serializable) containing the information needed to re-instantiate it.

property qlif_beta

class hgq.layers.snn.base.QSimpleSNN(*args, **kwargs)

Bases: _QSNN

Quantization-aware integrate-and-fire sequence layer.

get_config()

Returns the config of the object.

An object config is a Python dictionary (serializable) containing the information needed to re-instantiate it.

class hgq.layers.snn.base.QSimpleSNNCell(*args, **kwargs)

Bases: QLayerBaseSingleInput, SpikingNeuralCell

Quantization-aware integrate-and-fire RNN cell.

The membrane accumulator is not quantized by default. Input, emitted spikes, scalar parameters, and optionally recurrent state may be quantized.

build(input_shape)

call(inputs, states, training=None)

property enable_ebops

property enable_sq

fire(mem)

get_config()

Returns the config of the object.

An object config is a Python dictionary (serializable) containing the information needed to re-instantiate it.

property graded_spikes_factor_q

property qgraded_spikes_factor

reset(mem, spk)

property sq

class hgq.layers.snn.base.SpikingNeuralCell(*args, **kwargs)

Bases: Layer

Integrate-and-fire RNN cell.

build(input_shape)

call(inputs, states, training=None)

fire(mem)

classmethod from_config(config)

Creates an operation from its config.

This method is the reverse of get_config, capable of instantiating the same operation from the config dictionary.

Note: If you override this method, you might receive a serialized dtype config, which is a dict. You can deserialize it as follows:

```python if “dtype” in config and isinstance(config[“dtype”], dict):

policy = dtype_policies.deserialize(config[“dtype”])

```

Parameters:

config – A Python dictionary, typically the output of get_config.

Returns:

An operation instance.

get_config()

Returns the config of the object.

An object config is a Python dictionary (serializable) containing the information needed to re-instantiate it.

get_initial_state(batch_size=None)

reset(mem, spk)

reset_mechanisms = ('subtract', 'zero', 'none')

hgq.layers.snn.base.atan(alpha: float = 2.0)

Return an arc-tangent surrogate gradient callable.

Module contents

class hgq.layers.snn.ATan(alpha: float = 2.0)

Bases: object

Arc-tangent surrogate gradient for a Heaviside spike function.

get_config()

class hgq.layers.snn.LIFCell(*args, **kwargs)

Bases: SpikingNeuralCell

Leaky integrate-and-fire RNN cell.

build(input_shape)

call(inputs, states, training=None)

get_config()

Returns the config of the object.

An object config is a Python dictionary (serializable) containing the information needed to re-instantiate it.

class hgq.layers.snn.QLIF(*args, **kwargs)

Bases: _QSNN

Quantization-aware leaky integrate-and-fire sequence layer.

get_config()

Returns the config of the object.

An object config is a Python dictionary (serializable) containing the information needed to re-instantiate it.

class hgq.layers.snn.QLIFCell(*args, **kwargs)

Bases: QSimpleSNNCell

Quantization-aware leaky integrate-and-fire RNN cell.

property beta_q

build(input_shape)

call(inputs, states, training=None)

get_config()

Returns the config of the object.

An object config is a Python dictionary (serializable) containing the information needed to re-instantiate it.

property qlif_beta

class hgq.layers.snn.QSimpleSNN(*args, **kwargs)

Bases: _QSNN

Quantization-aware integrate-and-fire sequence layer.

get_config()

Returns the config of the object.

An object config is a Python dictionary (serializable) containing the information needed to re-instantiate it.

class hgq.layers.snn.QSimpleSNNCell(*args, **kwargs)

Bases: QLayerBaseSingleInput, SpikingNeuralCell

Quantization-aware integrate-and-fire RNN cell.

The membrane accumulator is not quantized by default. Input, emitted spikes, scalar parameters, and optionally recurrent state may be quantized.

build(input_shape)

call(inputs, states, training=None)

property enable_ebops

property enable_sq

fire(mem)

get_config()

Returns the config of the object.

An object config is a Python dictionary (serializable) containing the information needed to re-instantiate it.

property graded_spikes_factor_q

property qgraded_spikes_factor

reset(mem, spk)

property sq

class hgq.layers.snn.SpikingNeuralCell(*args, **kwargs)

Bases: Layer

Integrate-and-fire RNN cell.

build(input_shape)

call(inputs, states, training=None)

fire(mem)

classmethod from_config(config)

Creates an operation from its config.

This method is the reverse of get_config, capable of instantiating the same operation from the config dictionary.

Note: If you override this method, you might receive a serialized dtype config, which is a dict. You can deserialize it as follows:

```python if “dtype” in config and isinstance(config[“dtype”], dict):

policy = dtype_policies.deserialize(config[“dtype”])

```

Parameters:

config – A Python dictionary, typically the output of get_config.

Returns:

An operation instance.

get_config()

Returns the config of the object.

An object config is a Python dictionary (serializable) containing the information needed to re-instantiate it.

get_initial_state(batch_size=None)

reset(mem, spk)

reset_mechanisms = ('subtract', 'zero', 'none')

hgq.layers.snn.atan(alpha: float = 2.0)

Return an arc-tangent surrogate gradient callable.