def backward(ctx, grad_spike, grad_v_next):
grad_x, grad_v = _C_neuron.LIF_backward(grad_spike, grad_v_next,
ctx.saved_tensors[0],
ctx.saved_tensors[1],
ctx.reciprocal_tau,
ctx.detach_input)
return grad_x, grad_v, None, None, None, None, None, None, None
def LIF_backward(grad_spike: torch.Tensor, grad_v_next: torch.Tensor,
grad_s_to_h: torch.Tensor, grad_v_to_h: float,
reciprocal_tau: float, detach_input: bool):
'''
* :ref:`API in English <LIF_backward-en>`
.. _LIF_backward-cn:
:param reciprocal_tau: :math:`\\frac{1}{\\tau}`
:type reciprocal_tau: float
其余的参数参见 :ref:`backward_template <backward_template-cn>`。
梯度的计算按照
.. math::
\\frac{\\partial H_{t}}{\\partial X_{t}} & = \\frac{1}{\\tau}
\\frac{\\partial H_{t}}{\\partial V_{t-1}} & = 1 - \\frac{1}{\\tau}
* :ref:`中文API <LIF_backward-cn>`
.. _LIF_backward-en:
:param reciprocal_tau: :math:`\\frac{1}{\\tau}`
:type reciprocal_tau: float
See :ref:`backward_template <backward_template-en>` for more details about other args。
The gradients are calculated by
.. math::
\\frac{\\partial H_{t}}{\\partial X_{t}} & = \\frac{1}{\\tau}
\\frac{\\partial H_{t}}{\\partial V_{t-1}} & = 1 - \\frac{1}{\\tau}
'''
return _C_neuron.LIF_backward(grad_spike, grad_v_next, grad_s_to_h,
grad_v_to_h, reciprocal_tau, detach_input)