84fafd4599e29c68e63b90a38b56fcc65f289288,test/functions/test_inv_quad_log_det.py,TestInvQuadLogDetNonBatch,test_inv_quad_logdet_many_vectors_improper,#TestInvQuadLogDetNonBatch#,89

Before Change

    def test_inv_quad_logdet_many_vectors_improper(self):
        // Forward pass
        actual_inv_quad = self.mat_clone.inverse().matmul(self.vecs_clone).mul(self.vecs_clone).sum()
        actual_logdet = self.mat_clone.logdet()
        with gpytorch.settings.num_trace_samples(1000), gpytorch.settings.skip_logdet_forward(True):
            non_lazy_tsr = NonLazyTensor(self.mat)
            res_inv_quad, res_logdet = non_lazy_tsr.inv_quad_logdet(inv_quad_rhs=self.vecs, logdet=True)
        self.assertAlmostEqual(res_inv_quad.item(), actual_inv_quad.item(), places=1)
        self.assertAlmostEqual(res_logdet.item(), torch.zeros(non_lazy_tsr.batch_shape), places=1)

        // Backward
        actual_inv_quad.backward()
        actual_logdet.backward()
        res_inv_quad.backward(retain_graph=True)
        res_logdet.backward()

After Change

        rhs = torch.randn(*self.matrix_shape[:-1], 5)
        self._test_inv_quad_logdet(inv_quad_rhs=rhs, logdet=True)

    def test_inv_quad_logdet_many_vectors_improper(self):
        rhs = torch.randn(*self.matrix_shape[:-1], 5)
        self._test_inv_quad_logdet(inv_quad_rhs=rhs, logdet=True, improper_logdet=True)

    def test_inv_quad_only_many_vectors(self):
        rhs = torch.randn(*self.matrix_shape[:-1], 5)
        self._test_inv_quad_logdet(inv_quad_rhs=rhs, logdet=False)

Instances