Rename precision to dtype in spatialtf

theano/gpuarray/c_code/spatialtf_desc.c

@@ -27,7 +27,7 @@ int APPLY_SPECIFIC(spatialtf_desc)(cudnnSpatialTransformerDescriptor_t * desc,
     // Currently, only the bilinear sampler is supported by cuDNN,
     // so it is not available as a parameter
     err = cudnnSetSpatialTransformerNdDescriptor( *desc, CUDNN_SAMPLER_BILINEAR,
-        params->precision, params->nb_dims, out_tensor_dims );
+        params->dtype, params->nb_dims, out_tensor_dims );
     if ( CUDNN_STATUS_SUCCESS != err )
     {
         PyErr_Format( PyExc_MemoryError, 

theano/gpuarray/dnn.py

@@ -2840,9 +2840,9 @@ class GpuDnnSpatialTfDesc(COp):
     operations.
     """
 
-    __props__ = ('dimensions', 'precision')
+    __props__ = ('dimensions', 'dtype')
     params_type = ParamsType(nimages=int_t, nchannels=int_t, height=int_t, width=int_t,
-                             nb_dims=int_t, precision=cudnn.cudnnDataType_t)
+                             nb_dims=int_t, dtype=cudnn.cudnnDataType_t)
 
     def c_headers(self):
         return ['cudnn.h', 'cudnn_helper.h']
@@ -2859,7 +2859,7 @@ class GpuDnnSpatialTfDesc(COp):
     def do_constant_folding(self, node):
         return False
 
-    def __init__(self, dimensions, precision="float32"):
+    def __init__(self, dimensions, dtype="float32"):
         COp.__init__(self, ["c_code/spatialtf_desc.c"], "APPLY_SPECIFIC(spatialtf_desc)")
 
         # dimensions must have at least width and height
@@ -2871,8 +2871,8 @@ class GpuDnnSpatialTfDesc(COp):
         # not exceed 4 dimensions (width, height, num_feature_maps, num_images)
         assert len(self.dimensions) <= 4
 
-        assert cudnn.cudnnDataType_t.has_alias(precision)
-        self.precision = precision
+        assert cudnn.cudnnDataType_t.has_alias(dtype)
+        self.dtype = dtype
 
     def make_node(self):
         node = Apply(self, [],
@@ -2908,14 +2908,14 @@ class GpuDnnGridGenerator(DnnBase):
     operations.
     """
 
-    __props__ = ('precision',)
+    __props__ = ('dtype',)
     _cop_num_inputs = 3
     _cop_num_outputs = 1
 
-    def __init__(self, precision):
+    def __init__(self, dtype):
         DnnBase.__init__(self, ["c_code/spatialtf_grid.c"], "spatialtf_grid")
 
-        self.precision = precision
+        self.dtype = dtype
 
     def dnn_context(self, node):
         return node.outputs[0].type.context_name
@@ -2930,7 +2930,7 @@ class GpuDnnGridGenerator(DnnBase):
         assert theta.ndim == 3
 
         # Allocate GPU memory for grid of coordinates
-        grid = GpuArrayType(dtype=self.precision,
+        grid = GpuArrayType(dtype=self.dtype,
                             broadcastable=(False, False, False, False,),
                             context_name=context_name)()
 
@@ -2947,14 +2947,14 @@ class GpuDnnGridSampler(DnnBase):
     operations.
     """
 
-    __props__ = ('precision',)
+    __props__ = ('dtype',)
     _cop_num_inputs = 6
     _cop_num_outputs = 1
 
-    def __init__(self, precision):
+    def __init__(self, dtype):
         DnnBase.__init__(self, ["c_code/spatialtf_sampler.c"], "spatialtf_sampler")
 
-        self.precision = precision
+        self.dtype = dtype
 
     def dnn_context(self, node):
         return node.outputs[0].type.context_name
@@ -2967,7 +2967,7 @@ class GpuDnnGridSampler(DnnBase):
         grid = as_gpuarray_variable(grid, context_name)
         grid_dimensions = as_tensor_variable(grid_dimensions)
 
-        output = GpuArrayType(dtype=self.precision,
+        output = GpuArrayType(dtype=self.dtype,
                               broadcastable=img.type.ndim * (False,),
                               context_name=context_name)()
 
@@ -2993,7 +2993,7 @@ class GpuDnnGridSampler(DnnBase):
         pass
 
 
-def dnn_spatialtf(img, theta, grid_dims, alpha=None, beta=None, precision=None):
+def dnn_spatialtf(img, theta, grid_dims, alpha=None, beta=None, dtype=None):
     """
         GPU spatial transformer using cuDNN from NVIDIA.
     """
@@ -3001,18 +3001,18 @@ def dnn_spatialtf(img, theta, grid_dims, alpha=None, beta=None, precision=None):
     img = gpu_contiguous(img)
     theta = gpu_contiguous(theta)
 
-    precision = get_precision(precision, [img, theta])
+    dtype = get_precision(dtype, [img, theta])
 
     # Create spatial transformer descriptor
-    desc = GpuDnnSpatialTfDesc(grid_dims, precision)()
+    desc = GpuDnnSpatialTfDesc(grid_dims, dtype)()
 
     # Create grid dimensions variable
     grid_dims_var = as_tensor_variable(grid_dims)
 
     # Setup grid of coordinates
-    grid_coord = GpuDnnGridGenerator(precision)(grid_dims_var, theta, desc)
+    grid_coord = GpuDnnGridGenerator(dtype)(grid_dims_var, theta, desc)
 
-    grid_sampler = GpuDnnGridSampler(precision)(img, grid_coord, grid_dims_var, desc,
+    grid_sampler = GpuDnnGridSampler(dtype)(img, grid_coord, grid_dims_var, desc,
                                                 alpha, beta)
 
     return grid_sampler