Update torch_patch.py (#3174)

* Update torch_patch.py

Removed the PositiveDefinite() function for the one from PyTorch to be used instead, as is what less precise resulting in errors for certain cases.

* Update test_linear_models_eig.py

Passed scale_tril to dist.MultivariateNormal as keyword argument instead of positional argument as it otherwise gets identified as being a covariance matrix.

* Update gp.ipynb

* Update gp.ipynb

* Changed float precision to 64 in cell 27

pyro/distributions/torch_patch.py

@@ -71,17 +71,6 @@ def _HalfCauchy_logprob(self, value):
     return log_prob
 
 
-# TODO fix batch_shape have an extra singleton dimension upstream
-@patch_dependency("torch.distributions.constraints._PositiveDefinite.check")
-def _PositiveDefinite_check(self, value):
-    matrix_shape = value.shape[-2:]
-    batch_shape = value.shape[:-2]
-    flattened_value = value.reshape((-1,) + matrix_shape)
-    return torch.stack(
-        [torch.linalg.eigvalsh(v)[:1] > 0.0 for v in flattened_value]
-    ).view(batch_shape)
-
-
 @patch_dependency("torch.distributions.constraints._CorrCholesky.check")
 def _CorrCholesky_check(self, value):
     row_norm = torch.linalg.norm(value.detach(), dim=-1)

tests/contrib/oed/test_linear_models_eig.py

@@ -62,7 +62,7 @@ def marginal_guide(design, observation_labels, target_labels):
         torch.eye(3),
         constraint=torch.distributions.constraints.lower_cholesky,
     )
-    pyro.sample("y", dist.MultivariateNormal(mu, scale_tril))
+    pyro.sample("y", dist.MultivariateNormal(mu, scale_tril=scale_tril))
 
 
 def likelihood_guide(theta_dict, design, observation_labels, target_labels):

tutorial/source/gp.ipynb

@@ -70,7 +70,8 @@
     "\n",
     "smoke_test = \"CI\" in os.environ  # ignore; used to check code integrity in the Pyro repo\n",
     "assert pyro.__version__.startswith('1.8.4')\n",
-    "pyro.set_rng_seed(0)"
+    "pyro.set_rng_seed(0)\n",
+    "torch.set_default_tensor_type(torch.DoubleTensor)"
    ]
   },
   {
@@ -1054,7 +1055,7 @@
    "source": [
     "# only take petal length and petal width\n",
     "X = torch.from_numpy(\n",
-    "    df[df.columns[2:4]].values.astype(\"float32\"),\n",
+    "    df[df.columns[2:4]].values.astype(\"float64\"),\n",
     ")\n",
     "df[\"species\"] = df[\"species\"].astype(\"category\")\n",
     "# encode the species as 0, 1, 2\n",