Can't call numpy() on Tensor that requires grad. Use tensor.detach().numpy() instead.

[masanorihirano]

File ~/.local/lib/python3.8/site-packages/pytorch_extra_mhirano/experimental/variance_decomposition.py:272, in VarianceDecomposition.analysis_second_end(self)
    268 self.granger_causality_statistics = r * F
    269 # ToDo: torch.distributions.chi2.Chi2 does not support cdf at v1.11.0
    270 # chi2 = torch.distributions.chi2.Chi2(df=r)
    271 self.granger_causality_pvalues = 1 - torch.as_tensor(
--> 272     scipy.stats.chi2.cdf(self.granger_causality_statistics.cpu().numpy(), r)
    273 ).to(self.granger_causality_statistics)
    275 self.analysis_step = 0

RuntimeError: Can't call numpy() on Tensor that requires grad. Use tensor.detach().numpy() instead.