Add ConvTranspose1d module (pytorch#449)

docs/source/nn.rst

@@ -21,10 +21,13 @@ Convolution Layers
 .. autoclass:: Conv2d
     :members:
 
-.. autoclass:: ConvTranspose2d
+.. autoclass:: Conv3d
     :members:
 
-.. autoclass:: Conv3d
+.. autoclass:: ConvTranspose1d
+    :members:
+
+.. autoclass:: ConvTranspose2d
     :members:
 
 .. autoclass:: ConvTranspose3d

test/test_nn.py

@@ -962,7 +962,7 @@ def compare_cpu_gpu(outputs_cpu, outputs_gpu):
                 for bidirectional in (False, True):
                     for dropout in (0, 1): # Because of dropout randomness, can only compare 0 and 1
                         for batch_first in (False, True):
-                            num_directions = 2 if bidirectional else 1                            
+                            num_directions = 2 if bidirectional else 1
                             if batch_first:
                                 input_val = torch.randn(batch, seq_length, input_size)
                             else:
@@ -1211,6 +1211,19 @@ def add_test(test):
         input_size=(2, 4, 6),
         cudnn=True,
     ),
+    dict(
+        module_name='ConvTranspose1d',
+        constructor_args=(3, 4, 3, (3,), 1, (1,)),
+        cudnn=True,
+        input_size=(1, 3, 7)
+    ),
+    dict(
+        module_name='ConvTranspose1d',
+        constructor_args=(3, 4, 3, 2, 1, 1, 1, False),
+        input_size=(1, 3, 6),
+        cudnn=True,
+        desc='no_bias'
+    ),
     dict(
         module_name='MaxPool1d',
         constructor_args=(4,),

torch/nn/functional.py

@@ -124,6 +124,13 @@ def conv3d(input, weight, bias=None, stride=1, padding=0, dilation=1,
     return f(input, weight, bias) if bias is not None else f(input, weight)
 
 
+def conv_transpose1d(input, weight, bias=None, stride=1, padding=0,
+                     output_padding=0, groups=1):
+    f = ConvNd(_single(stride), _single(padding), _single(1), True,
+               _single(output_padding), groups)
+    return f(input, weight, bias) if bias is not None else f(input, weight)
+
+
 def conv_transpose2d(input, weight, bias=None, stride=1, padding=0,
                      output_padding=0, groups=1):
     """Applies a 2D transposed convolution operator over an input image

torch/nn/modules/__init__.py

@@ -1,6 +1,7 @@
 from .module import Module
 from .linear import Linear
-from .conv import Conv1d, Conv2d, ConvTranspose2d, Conv3d, ConvTranspose3d
+from .conv import Conv1d, Conv2d, Conv3d, \
+    ConvTranspose1d, ConvTranspose2d, ConvTranspose3d
 from .activation import Threshold, ReLU, Hardtanh, ReLU6, Sigmoid, Tanh, \
     Softmax, Softmax2d, LogSoftmax, ELU, Hardshrink, LeakyReLU, LogSigmoid, \
     Softplus, Softshrink, PReLU, Softsign, Softmin, Tanhshrink, RReLU