change sigmoid and tanh to torch from nn.functional

dhpollack · dhpollack · commit 5be878992e2e · 2018-08-08T10:39:31.000+02:00
diff --git a/python/lltm.py b/python/lltm.py
@@ -31,14 +31,14 @@ def forward(self, input, state):
         # Split the combined gate weight matrix into its components.
         gates = gate_weights.chunk(3, dim=1)
 
-        input_gate = F.sigmoid(gates[0])
-        output_gate = F.sigmoid(gates[1])
+        input_gate = torch.sigmoid(gates[0])
+        output_gate = torch.sigmoid(gates[1])
         # Here we use an ELU instead of the usual tanh.
         candidate_cell = F.elu(gates[2])
 
         # Compute the new cell state.
         new_cell = old_cell + candidate_cell * input_gate
         # Compute the new hidden state and output.
-        new_h = F.tanh(new_cell) * output_gate
+        new_h = torch.tanh(new_cell) * output_gate
 
         return new_h, new_cell
diff --git a/python/lltm_baseline.py b/python/lltm_baseline.py
@@ -9,12 +9,12 @@
 
 
 def d_sigmoid(z):
-    s = F.sigmoid(z)
+    s = torch.sigmoid(z)
     return (1 - s) * s
 
 
 def d_tanh(z):
-    t = F.tanh(z)
+    t = torch.tanh(z)
     return 1 - (t * t)
 
 
@@ -32,12 +32,12 @@ def forward(ctx, input, weights, bias, old_h, old_cell):
         gate_weights = F.linear(X, weights, bias)
         gates = gate_weights.chunk(3, dim=1)
 
-        input_gate = F.sigmoid(gates[0])
-        output_gate = F.sigmoid(gates[1])
+        input_gate = torch.sigmoid(gates[0])
+        output_gate = torch.sigmoid(gates[1])
         candidate_cell = F.elu(gates[2])
 
         new_cell = old_cell + candidate_cell * input_gate
-        new_h = F.tanh(new_cell) * output_gate
+        new_h = torch.tanh(new_cell) * output_gate
 
         ctx.save_for_backward(X, weights, input_gate, output_gate, old_cell,
                               new_cell, candidate_cell, gate_weights)
@@ -51,7 +51,7 @@ def backward(ctx, grad_h, grad_cell):
 
         d_input = d_weights = d_bias = d_old_h = d_old_cell = None
 
-        d_output_gate = F.tanh(new_cell) * grad_h
+        d_output_gate = torch.tanh(new_cell) * grad_h
         d_tanh_new_cell = output_gate * grad_h
         d_new_cell = d_tanh(new_cell) * d_tanh_new_cell + grad_cell
 
