7391aec984c1e9e0899d51e6b0538483c37ec8fb,torchdiffeq/_impl/adjoint.py,,odeint_adjoint,#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#,115

Before Change

        adjoint_options = options

    tensor_input = False
    if torch.is_tensor(y0):

        class TupleFunc(nn.Module):

            def __init__(self, base_func):
                super(TupleFunc, self).__init__()
                self.base_func = base_func

            def forward(self, t, y):
                return (self.base_func(t, y[0]),)

        tensor_input = True
        y0 = (y0,)
        func = TupleFunc(func)

    params = tuple(func.parameters())
    if adjoint_buffers:
        params = params + tuple(buffer for buffer in func.buffers() if buffer.requires_grad)
    n_tensors = len(y0)

    ys = OdeintAdjointMethod.apply(func, t, rtol, atol, method, options, adjoint_rtol, adjoint_atol, adjoint_method,
                                   adjoint_options, n_tensors, *params, *y0)

After Change

    if adjoint_params is None:
        adjoint_params = tuple(func.parameters())

    tensor_input, shapes, func, y0, t, options = _check_inputs(func, y0, t, options)

    solution = OdeintAdjointMethod.apply(func, y0, t, rtol, atol, method, options, adjoint_rtol, adjoint_atol,
                                         adjoint_method, adjoint_options, t.requires_grad, *adjoint_params)

Instances