gpytorch/variational/_variational_strategy.py

Survived

Mutant 184

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -16,7 +16,7 @@
     Abstract base class for all Variational Strategies.
     """
 
-    def __init__(self, model, inducing_points, variational_distribution, learn_inducing_locations=True):
+    def __init__(self, model, inducing_points, variational_distribution, learn_inducing_locations=False):
         super().__init__()
 
         # Model

Mutant 189

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -25,7 +25,7 @@
         # Inducing points
         inducing_points = inducing_points.clone()
         if inducing_points.dim() == 1:
-            inducing_points = inducing_points.unsqueeze(-1)
+            inducing_points = inducing_points.unsqueeze(+1)
         if learn_inducing_locations:
             self.register_parameter(name="inducing_points", parameter=torch.nn.Parameter(inducing_points))
         else:

Mutant 196

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -33,7 +33,7 @@
 
         # Variational distribution
         self._variational_distribution = variational_distribution
-        self.register_buffer("variational_params_initialized", torch.tensor(0))
+        self.register_buffer("variational_params_initialized", torch.tensor(1))
 
     def _expand_inputs(self, x, inducing_points):
         """

Mutant 198

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -39,7 +39,7 @@
         """
         Pre-processing step in __call__ to make x the same batch_shape as the inducing points
         """
-        batch_shape = _mul_broadcast_shape(inducing_points.shape[:-2], x.shape[:-2])
+        batch_shape = _mul_broadcast_shape(inducing_points.shape[:-3], x.shape[:-2])
         inducing_points = inducing_points.expand(*batch_shape, *inducing_points.shape[-2:])
         x = x.expand(*batch_shape, *x.shape[-2:])
         return x, inducing_points

Mutant 203

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -40,7 +40,7 @@
         Pre-processing step in __call__ to make x the same batch_shape as the inducing points
         """
         batch_shape = _mul_broadcast_shape(inducing_points.shape[:-2], x.shape[:-2])
-        inducing_points = inducing_points.expand(*batch_shape, *inducing_points.shape[-2:])
+        inducing_points = inducing_points.expand(*batch_shape, *inducing_points.shape[-3:])
         x = x.expand(*batch_shape, *x.shape[-2:])
         return x, inducing_points
 

Mutant 208

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -44,7 +44,6 @@
         x = x.expand(*batch_shape, *x.shape[-2:])
         return x, inducing_points
 
-    @abstractproperty
     @cached(name="prior_distribution_memo")
     def prior_distribution(self):
         r"""

Mutant 209

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -45,7 +45,7 @@
         return x, inducing_points
 
     @abstractproperty
-    @cached(name="prior_distribution_memo")
+    @cached(name="XXprior_distribution_memoXX")
     def prior_distribution(self):
         r"""
         The :func:`~gpytorch.variational.VariationalStrategy.prior_distribution` method determines how to compute the

Mutant 210

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -45,7 +45,7 @@
         return x, inducing_points
 
     @abstractproperty
-    @cached(name="prior_distribution_memo")
+
     def prior_distribution(self):
         r"""
         The :func:`~gpytorch.variational.VariationalStrategy.prior_distribution` method determines how to compute the

Mutant 212

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -58,7 +58,7 @@
         raise NotImplementedError
 
     @property
-    @cached(name="variational_distribution_memo")
+    @cached(name="XXvariational_distribution_memoXX")
     def variational_distribution(self):
         return self._variational_distribution()
 

Mutant 213

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -58,7 +58,7 @@
         raise NotImplementedError
 
     @property
-    @cached(name="variational_distribution_memo")
+
     def variational_distribution(self):
         return self._variational_distribution()
 

Mutant 214

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -90,7 +90,7 @@
 
         :rtype: torch.Tensor
         """
-        with settings.max_preconditioner_size(0):
+        with settings.max_preconditioner_size(1):
             kl_divergence = torch.distributions.kl.kl_divergence(self.variational_distribution, self.prior_distribution)
         return kl_divergence
 

Mutant 216

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -94,7 +94,7 @@
             kl_divergence = torch.distributions.kl.kl_divergence(self.variational_distribution, self.prior_distribution)
         return kl_divergence
 
-    def train(self, mode=True):
+    def train(self, mode=False):
         # Make sure we are clearing the cache if we change modes
         if (self.training and not mode) or mode:
             clear_cache_hook(self)

Mutant 217

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -96,7 +96,7 @@
 
     def train(self, mode=True):
         # Make sure we are clearing the cache if we change modes
-        if (self.training and not mode) or mode:
+        if (self.training and  mode) or mode:
             clear_cache_hook(self)
         return super().train(mode=mode)
 

Mutant 218

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -96,7 +96,7 @@
 
     def train(self, mode=True):
         # Make sure we are clearing the cache if we change modes
-        if (self.training and not mode) or mode:
+        if (self.training or not mode) or mode:
             clear_cache_hook(self)
         return super().train(mode=mode)
 

Mutant 219

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -96,7 +96,7 @@
 
     def train(self, mode=True):
         # Make sure we are clearing the cache if we change modes
-        if (self.training and not mode) or mode:
+        if (self.training and not mode) and mode:
             clear_cache_hook(self)
         return super().train(mode=mode)
 

Mutant 221

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -100,7 +100,7 @@
             clear_cache_hook(self)
         return super().train(mode=mode)
 
-    def __call__(self, x, prior=False):
+    def __call__(self, x, prior=True):
         # If we're in prior mode, then we're done!
         if prior:
             return self.model.forward(x)

Mutant 223

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -109,7 +109,7 @@
         if self.training:
             clear_cache_hook(self)
         # (Maybe) initialize variational distribution
-        if not self.variational_params_initialized.item():
+        if  self.variational_params_initialized.item():
             prior_dist = self.prior_distribution
             self._variational_distribution.initialize_variational_distribution(prior_dist)
             self.variational_params_initialized.fill_(1)

Mutant 225

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -112,7 +112,7 @@
         if not self.variational_params_initialized.item():
             prior_dist = self.prior_distribution
             self._variational_distribution.initialize_variational_distribution(prior_dist)
-            self.variational_params_initialized.fill_(1)
+            self.variational_params_initialized.fill_(2)
 
         # Ensure inducing_points and x are the same size
         inducing_points = self.inducing_points

Mutant 227

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -116,7 +116,7 @@
 
         # Ensure inducing_points and x are the same size
         inducing_points = self.inducing_points
-        if inducing_points.shape[:-2] != x.shape[:-2]:
+        if inducing_points.shape[:+2] != x.shape[:-2]:
             x, inducing_points = self._expand_inputs(x, inducing_points)
 
         # Get p(u)/q(u)

Mutant 228

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -116,7 +116,7 @@
 
         # Ensure inducing_points and x are the same size
         inducing_points = self.inducing_points
-        if inducing_points.shape[:-2] != x.shape[:-2]:
+        if inducing_points.shape[:-3] != x.shape[:-2]:
             x, inducing_points = self._expand_inputs(x, inducing_points)
 
         # Get p(u)/q(u)

Mutant 230

--- gpytorch/variational/_variational_strategy.py
+++ gpytorch/variational/_variational_strategy.py
@@ -116,7 +116,7 @@
 
         # Ensure inducing_points and x are the same size
         inducing_points = self.inducing_points
-        if inducing_points.shape[:-2] != x.shape[:-2]:
+        if inducing_points.shape[:-2] != x.shape[:+2]:
             x, inducing_points = self._expand_inputs(x, inducing_points)
 
         # Get p(u)/q(u)