qualia_plugin_snn.learningmodel.pytorch.layers package

Subpackages

• qualia_plugin_snn.learningmodel.pytorch.layers.spikingjelly package
  • Submodules
    • qualia_plugin_snn.learningmodel.pytorch.layers.spikingjelly.Add module
      • Add
    • qualia_plugin_snn.learningmodel.pytorch.layers.spikingjelly.QuantizedAdd module
      • QuantizedAdd
    • qualia_plugin_snn.learningmodel.pytorch.layers.spikingjelly.layers1d module
      • AvgPool
      • BatchNorm
      • Conv
      • MaxPool
      • QuantizedBatchNorm
      • QuantizedConv
      • QuantizedMaxPool
    • qualia_plugin_snn.learningmodel.pytorch.layers.spikingjelly.layers2d module
      • AvgPool
      • BatchNorm
      • Conv
      • MaxPool
      • QuantizedBatchNorm
      • QuantizedConv
      • QuantizedMaxPool
    • qualia_plugin_snn.learningmodel.pytorch.layers.spikingjelly.quantized_layers module
      • QuantizedLinear
    • qualia_plugin_snn.learningmodel.pytorch.layers.spikingjelly.quantized_layers1d module
      • QuantizedConv1d
      • QuantizedMaxPool1d
      • QuantizedBatchNorm1d
    • qualia_plugin_snn.learningmodel.pytorch.layers.spikingjelly.quantized_layers2d module
      • QuantizedConv2d
      • QuantizedMaxPool2d
      • QuantizedBatchNorm2d
  • Module contents

Submodules

• qualia_plugin_snn.learningmodel.pytorch.layers.CustomNode module
  • heaviside()
  • SpikeFunctionSigmoid
    • SpikeFunctionSigmoid.forward()
    • SpikeFunctionSigmoid.backward()
    • SpikeFunctionSigmoid.apply()
  • IFSRL
    • IFSRL.__init__()
    • IFSRL.v
    • IFSRL.get_coeffs()
    • IFSRL.set_coeffs()
    • IFSRL.reset()
    • IFSRL.ifsrl_fn()
    • IFSRL.forward()
    • IFSRL.training
  • ATIF
    • ATIF.__init__()
    • ATIF.v
    • ATIF.supported_backends
    • ATIF.get_coeffs()
    • ATIF.set_coeffs()
    • ATIF.training
    • ATIF.ifsrl_fn()
    • ATIF.single_step_forward()
• qualia_plugin_snn.learningmodel.pytorch.layers.quantized_SNN_layers module
  • QuantizedLIFNode
    • QuantizedLIFNode.v
    • QuantizedLIFNode.v_threshold
    • QuantizedLIFNode.v_reset
    • QuantizedLIFNode.tau
    • QuantizedLIFNode.__init__()
    • QuantizedLIFNode.supported_backends
    • QuantizedLIFNode.neuronal_charge()
    • QuantizedLIFNode.single_step_forward()
    • QuantizedLIFNode.multi_step_forward()
    • QuantizedLIFNode.reciprocal_tau
    • QuantizedLIFNode.quantize_v_and_hyperparams()
    • QuantizedLIFNode.get_hyperparams_tensor()
    • QuantizedLIFNode.weights_q
    • QuantizedLIFNode.weights_round_mode
  • QuantizedIFNode
    • QuantizedIFNode.v
    • QuantizedIFNode.v_threshold
    • QuantizedIFNode.v_reset
    • QuantizedIFNode.__init__()
    • QuantizedIFNode.supported_backends
    • QuantizedIFNode.neuronal_charge()
    • QuantizedIFNode.single_step_forward()
    • QuantizedIFNode.multi_step_forward()
    • QuantizedIFNode.quantize_v_and_hyperparams()
    • QuantizedIFNode.get_hyperparams_tensor()
    • QuantizedIFNode.weights_q
    • QuantizedIFNode.weights_round_mode
  • QuantizedATIF
    • QuantizedATIF.__init__()
    • QuantizedATIF.ifsrl_fn()
    • QuantizedATIF.get_coeffs()
    • QuantizedATIF.set_coeffs()
    • QuantizedATIF.weights_q
    • QuantizedATIF.weights_round_mode

Module contents

Contains implementation for custom or quantized layers for spiking neural networks.

class qualia_plugin_snn.learningmodel.pytorch.layers.ATIF

Bases: BaseNode

IFSRLSJ: Integrate and Fire soft-reset with learnable Vth and activation scaling, based on spikingjelly.

__init__(v_threshold: float = 1.0, vth_init_l: float = 0.8, vth_init_h: float = 1.0, alpha: float = 1.0, device: str = 'cpu') → None

Construct ATIF.

Parameters:

• v_threshold (float) – Factor to apply to the uniform initialization bounds

• vth_init_l (float) – Lower bound for uniform initialization of threshold Tensor

• vth_init_h (float) – Higher bound for uniform initialization of threshold Tensor

• alpha (float) – Sigmoig surrogate scale factor

• device (str) – Device to run the computation on

Return type:

None

v: torch.Tensor

property supported_backends: tuple[Literal['torch']]

Supported step_mode and backend.

Only single-step mode with torch backend is supported.

Returns:

Tuple of 'torch' if step_mode is 's'

Raises:

ValueError – When step_mode is not 's'

get_coeffs() → Tensor

Return the Tensor of threshold v_threshold.

Returns:

Tensor of threshold v_threshold

Return type:

Tensor

set_coeffs(v_threshold: Tensor) → None

Replace the Tensor of threshold v_threshold.

Parameters:

v_threshold (Tensor) – New Tensor of threshold to replace v_threshold

Return type:

None

ifsrl_fn(x: Tensor) → Tensor

Integrate-and-Fire soft-reset neuron with learnable threshold.

Parameters:

x (Tensor) – Input tensor

Returns:

Output tensor

Return type:

Tensor

single_step_forward(x: Tensor) → Tensor

Single-step mode forward of ATIF.

Calls ifsrl_fn().

Parameters:

x (Tensor) – Input tensor

Returns:

Output tensor

Return type:

Tensor

class qualia_plugin_snn.learningmodel.pytorch.layers.IFSRL

Bases: Module

IFSRL: Integrate and Fire soft-reset with learnable Vth and activation scaling.

__init__(v_threshold: float = 1.0, vth_init_l: float = 0.8, vth_init_h: float = 1.0, alpha: float = 1.0, device: str = 'cpu') → None

Construct IFSRL.

Parameters:

• v_threshold (float) – Factor to apply to the uniform initialization bounds

• vth_init_l (float) – Lower bound for uniform initialization of threshold Tensor

• vth_init_h (float) – Higher bound for uniform initialization of threshold Tensor

• alpha (float) – Sigmoig surrogate scale factor

• device (str) – Device to run the computation on

Return type:

None

v: torch.Tensor

get_coeffs() → Tensor

Return the Tensor of threshold vp_th.

Returns:

Tensor of threshold vp_th

Return type:

Tensor

set_coeffs(vp_th: Tensor) → None

Replace the Tensor of threshold vp_th.

Parameters:

vp_th (Tensor) – New Tensor of threshold to replace vp_th

Return type:

None

reset() → None

Reset potential to 0.

Return type:

None

ifsrl_fn(x: Tensor) → Tensor

Integrate-and-Fire soft-reset neuron with learnable threshold.

Parameters:

x (Tensor) – Input tensor

Returns:

Output tensor

Return type:

Tensor

forward(input: Tensor) → Tensor

Forward of ifsrl_fn().

Parameters:

input (Tensor) – Input tensor

Returns:

Output tensor

Return type:

Tensor