Add checks for input dimensions in GpuDnnTrasnformerGradI

theano/gpuarray/dnn.py

@@ -2937,17 +2937,17 @@ class GpuDnnTransformer(DnnBase):
         return Apply(self, inputs, outputs)
 
     def L_op(self, inputs, outputs, grads):
-        img, theta, grid_dims, desc, alpha, beta = inputs
+        img, theta, output, desc, alpha, beta = inputs
 
         _, grid = outputs
         dy = grads[0]
 
-        dimg, dgrid = GpuDnnTransformerGradI(self.dtype)(img, theta, grid, grid_dims, dy,
+        dimg, dgrid = GpuDnnTransformerGradI(self.dtype)(img, theta, grid, dy,
                                                          desc, alpha, beta)
         dtheta = GpuDnnTransformerGradT(self.dtype)(dgrid, desc)
 
         return [dimg, dtheta,
-                theano.gradient.grad_undefined(self, 2, grid_dims),
+                theano.gradient.grad_undefined(self, 2, output),
                 DisconnectedType()(),
                 theano.gradient.grad_undefined(self, 4, alpha),
                 theano.gradient.grad_undefined(self, 5, beta)]
@@ -2971,21 +2971,31 @@ class GpuDnnTransformerGradI(DnnBase):
         self.dtype = dtype
 
     def make_node(self, img, theta, grid, dy, desc, alpha, beta):
-        context_name = infer_context_name(img)
+        context_name = infer_context_name(img, theta, grid, dy, desc)
 
-        if img.ndim != 4:
-            raise RuntimeError('img must have 4 dimensions.')
-        if theta.ndim != 3:
-            raise RuntimeError('theta must have 3 dimensions')
+        if (not isinstance(desc.type, CDataType) or
+                desc.type.ctype != 'cudnnSpatialTransformerDescriptor_t'):
+            raise ValueError('desc must be cudnnSpatialTransformerDescriptor_t')
 
         img = as_gpuarray_variable(gpu_contiguous(img), context_name)
+        if img.ndim != 4:
+            raise TypeError('img must have 4 dimensions.')
+
         theta = as_gpuarray_variable(gpu_contiguous(theta), context_name)
+        if theta.ndim != 3:
+            raise TypeError('theta must have 3 dimensions')
+
         grid = as_gpuarray_variable(gpu_contiguous(grid), context_name)
+        if img.ndim != grid.ndim:
+            raise TypeError('grid should have the same number of dimensions as img')
 
         # Setup grid dimensions from descriptor's input
         grid_dims = as_tensor_variable(desc.owner.inputs[0])
 
         dy = as_gpuarray_variable(dy, context_name)
+        if img.ndim != 4:
+            raise TypeError('img must have 4 dimensions.')
+
         alpha = as_scalar(alpha)
         beta = as_scalar(beta)