pytorch · weiyangfb · Jun 21, 2018 · Jun 21, 2018 · Jun 21, 2018 · Jun 21, 2018
diff --git a/test/test_jit.py b/test/test_jit.py
@@ -70,10 +70,10 @@ def LSTMCell(input, hidden, w_ih, w_hh, b_ih=None, b_hh=None):
     gates = F.linear(input, w_ih, b_ih) + F.linear(hx, w_hh, b_hh)
 
     ingate, forgetgate, cellgate, outgate = gates.chunk(4, 1)
-    ingate = F.sigmoid(ingate)
-    forgetgate = F.sigmoid(forgetgate)
+    ingate = torch.sigmoid(ingate)
+    forgetgate = torch.sigmoid(forgetgate)
     cellgate = F.tanh(cellgate)
-    outgate = F.sigmoid(outgate)
+    outgate = torch.sigmoid(outgate)
 
     cy = (forgetgate * cx) + (ingate * cellgate)
     hy = outgate * F.tanh(cy)
@@ -4476,7 +4476,7 @@ def reparameterize(self, mu, logvar):
 
             def decode(self, z):
                 h3 = F.relu(self.fc3(z))
-                return F.sigmoid(self.fc4(h3))
+                return torch.sigmoid(self.fc4(h3))
 
             def forward(self, x):
                 mu, logvar = self.encode(x.view(-1, 784))

diff --git a/torch/distributions/transforms.py b/torch/distributions/transforms.py
@@ -6,7 +6,7 @@
 from torch.distributions import constraints
 from torch.distributions.utils import (_sum_rightmost, broadcast_all,
                                        lazy_property)
-from torch.nn.functional import pad, sigmoid
+from torch.nn.functional import pad
 
 __all__ = [
     'AbsTransform',
@@ -341,7 +341,7 @@ def __eq__(self, other):
         return isinstance(other, SigmoidTransform)
 
     def _call(self, x):
-        return sigmoid(x)
+        return torch.sigmoid(x)
 
     def _inverse(self, y):
         return y.log() - (-y).log1p()
@@ -483,7 +483,7 @@ def __eq__(self, other):
 
     def _call(self, x):
         offset = (x.shape[-1] + 1) - x.new([1]).expand(x.shape).cumsum(-1)
-        z = sigmoid(x - offset.log())
+        z = torch.sigmoid(x - offset.log())
         z_cumprod = (1 - z).cumprod(-1)
         y = pad(z, (0, 1), value=1) * pad(z_cumprod, (1, 0), value=1)
         return y
@@ -497,7 +497,7 @@ def _inverse(self, y):
 
     def log_abs_det_jacobian(self, x, y):
         offset = (x.shape[-1] + 1) - x.new([1]).expand(x.shape).cumsum(-1)
-        z = sigmoid(x - offset.log())
+        z = torch.sigmoid(x - offset.log())
         detJ = ((1 - z).log() + y[..., :-1].log()).sum(-1)
         return detJ
 

diff --git a/torch/distributions/utils.py b/torch/distributions/utils.py
@@ -116,7 +116,7 @@ def logits_to_probs(logits, is_binary=False):
     the log probabilities (possibly unnormalized) of the events.
     """
     if is_binary:
-        return F.sigmoid(logits)
+        return torch.sigmoid(logits)
     return F.softmax(logits, dim=-1)
 
 

diff --git a/torch/nn/_functions/rnn.py b/torch/nn/_functions/rnn.py
@@ -2,6 +2,7 @@
 from torch.autograd import NestedIOFunction
 import torch.backends.cudnn as cudnn
 from .. import functional as F
+import torch
 from .thnn import rnnFusedPointwise as fusedBackend
 import itertools
 from functools import partial
@@ -18,7 +19,7 @@ def RNNReLUCell(input, hidden, w_ih, w_hh, b_ih=None, b_hh=None):
 
 
 def RNNTanhCell(input, hidden, w_ih, w_hh, b_ih=None, b_hh=None):
-    hy = F.tanh(F.linear(input, w_ih, b_ih) + F.linear(hidden, w_hh, b_hh))
+    hy = torch.tanh(F.linear(input, w_ih, b_ih) + F.linear(hidden, w_hh, b_hh))
     return hy
 
 
@@ -34,13 +35,13 @@ def LSTMCell(input, hidden, w_ih, w_hh, b_ih=None, b_hh=None):
 
     ingate, forgetgate, cellgate, outgate = gates.chunk(4, 1)
 
-    ingate = F.sigmoid(ingate)
-    forgetgate = F.sigmoid(forgetgate)
-    cellgate = F.tanh(cellgate)
-    outgate = F.sigmoid(outgate)
+    ingate = torch.sigmoid(ingate)
+    forgetgate = torch.sigmoid(forgetgate)
+    cellgate = torch.tanh(cellgate)
+    outgate = torch.sigmoid(outgate)
 
     cy = (forgetgate * cx) + (ingate * cellgate)
-    hy = outgate * F.tanh(cy)
+    hy = outgate * torch.tanh(cy)
 
     return hy, cy
 
@@ -58,9 +59,9 @@ def GRUCell(input, hidden, w_ih, w_hh, b_ih=None, b_hh=None):
     i_r, i_i, i_n = gi.chunk(3, 1)
     h_r, h_i, h_n = gh.chunk(3, 1)
 
-    resetgate = F.sigmoid(i_r + h_r)
-    inputgate = F.sigmoid(i_i + h_i)
-    newgate = F.tanh(i_n + resetgate * h_n)
+    resetgate = torch.sigmoid(i_r + h_r)
+    inputgate = torch.sigmoid(i_i + h_i)
+    newgate = torch.tanh(i_n + resetgate * h_n)
     hy = newgate + inputgate * (hidden - newgate)
 
     return hy

diff --git a/torch/nn/functional.py b/torch/nn/functional.py
@@ -1009,6 +1009,7 @@ def tanh(input):
 
     See :class:`~torch.nn.Tanh` for more details.
     """
+    warnings.warn("nn.functional.tanh is deprecated. Use torch.tanh instead.")
     return input.tanh()
 
 
@@ -1019,11 +1020,10 @@ def sigmoid(input):
 
     See :class:`~torch.nn.Sigmoid` for more details.
     """
+    warnings.warn("nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.")
     return input.sigmoid()
 
 
-# etc.
-
 def linear(input, weight, bias=None):
     r"""
     Applies a linear transformation to the incoming data: :math:`y = xA^T + b`.