Replace from torch.nn.functional import ... with `import torch.nn.f…

…unctional as F`

These two expressions are almost identical, but it turns out that the
latter is (currently) necessary to visualize models using torchlens.
See johnmarktaylor91/torchlens#18.  While this is a pretty minor
benefit, it's also a pretty minor change.

Signed-off-by: Kale Kundert <[email protected]>

escnn/nn/modules/conv/r2_transposed_convolution.py

@@ -1,5 +1,5 @@
 
-from torch.nn.functional import conv_transpose2d
+import torch.nn.functional as F
 
 import escnn.nn
 from escnn.nn import FieldType
@@ -127,7 +127,7 @@ def forward(self, input: GeometricTensor):
             _filter, _bias = self.expand_parameters()
 
         # use it for convolution and return the result
-        output = conv_transpose2d(
+        output = F.conv_transpose2d(
                         input.tensor, _filter,
                         padding=self.padding,
                         output_padding=self.output_padding,

escnn/nn/modules/conv/r2convolution.py

@@ -1,4 +1,4 @@
-from torch.nn.functional import conv2d, pad
+import torch.nn.functional as F
 
 from escnn.nn import FieldType
 from escnn.nn import GeometricTensor
@@ -214,14 +214,14 @@ def forward(self, input: GeometricTensor):
         # use it for convolution and return the result
 
         if self.padding_mode == 'zeros':
-            output = conv2d(input.tensor, _filter,
+            output = F.conv2d(input.tensor, _filter,
                             stride=self.stride,
                             padding=self.padding,
                             dilation=self.dilation,
                             groups=self.groups,
                             bias=_bias)
         else:
-            output = conv2d(pad(input.tensor, self._reversed_padding_repeated_twice, self.padding_mode),
+            output = F.conv2d(F.pad(input.tensor, self._reversed_padding_repeated_twice, self.padding_mode),
                             _filter,
                             stride=self.stride,
                             dilation=self.dilation,

escnn/nn/modules/conv/r3_transposed_convolution.py

@@ -1,6 +1,6 @@
 import gc
 
-from torch.nn.functional import conv_transpose3d
+import torch.nn.functional as F
 
 import escnn.nn
 from escnn.nn import FieldType
@@ -128,7 +128,7 @@ def forward(self, input: GeometricTensor):
 
 
         # use it for convolution and return the result
-        output = conv_transpose3d(
+        output = F.conv_transpose3d(
                         input.tensor, _filter,
                         padding=self.padding,
                         output_padding=self.output_padding,

escnn/nn/modules/conv/r3convolution.py

@@ -1,4 +1,4 @@
-from torch.nn.functional import conv3d, pad
+import torch.nn.functional as F
 
 import escnn.nn
 from escnn.nn import FieldType
@@ -207,14 +207,14 @@ def forward(self, input: GeometricTensor):
 
         # use it for convolution and return the result
         if self.padding_mode == 'zeros':
-            output = conv3d(input.tensor, _filter,
+            output = F.conv3d(input.tensor, _filter,
                             stride=self.stride,
                             padding=self.padding,
                             dilation=self.dilation,
                             groups=self.groups,
                             bias=_bias)
         else:
-            output = conv3d(pad(input.tensor, self._reversed_padding_repeated_twice, self.padding_mode),
+            output = F.conv3d(F.pad(input.tensor, self._reversed_padding_repeated_twice, self.padding_mode),
                             _filter,
                             stride=self.stride,
                             dilation=self.dilation,

escnn/nn/modules/linear.py

@@ -9,7 +9,7 @@
 from escnn.nn.modules.basismanager import BlocksBasisExpansion
 
 from torch.nn import Parameter
-from torch.nn.functional import linear
+import torch.nn.functional as F
 import torch
 import numpy as np
 
@@ -202,7 +202,7 @@ def forward(self, input: GeometricTensor):
             # retrieve the matrix and the bias
             _matrix, _bias = self.expand_parameters()
 
-        output = linear(input.tensor, _matrix, bias=_bias)
+        output = F.linear(input.tensor, _matrix, bias=_bias)
 
         return GeometricTensor(output, self.out_type, input.coords)
 

escnn/nn/modules/nonlinearities/tensor.py

@@ -12,7 +12,6 @@
 from escnn.nn.modules.basismanager import BlocksBasisExpansion
 
 from torch.nn import Parameter
-from torch.nn.functional import linear
 
 from typing import List, Tuple, Any
 

escnn/nn/modules/rdupsampling.py

@@ -15,7 +15,7 @@
 
 import math
 
-from torch.nn.functional import interpolate
+import torch.nn.functional as F
 
 __all__ = ["R2Upsampling", "R3Upsampling"]
 
@@ -105,12 +105,12 @@ def forward(self, input: GeometricTensor):
         assert len(input.shape) == 2 + self.d, (input.shape, self.d)
 
         if self._align_corners is None:
-            output = interpolate(input.tensor,
+            output = F.interpolate(input.tensor,
                                  size=self._size,
                                  scale_factor=self._scale_factor,
                                  mode=self._mode)
         else:
-            output = interpolate(input.tensor,
+            output = F.interpolate(input.tensor,
                                  size=self._size,
                                  scale_factor=self._scale_factor,
                                  mode=self._mode,
@@ -413,4 +413,4 @@ def check_equivariance(self, atol: float = 0.1, rtol: float = 0.1):
 
             errors.append((el, errs.mean()))
 
-        return errors
+        return errors