bug fix

utils/generic_utils.py

@@ -133,7 +133,8 @@ def lr_decay(init_lr, global_step, warmup_steps):
 
 
 def create_attn_mask(N, T, g=0.05):
-    r'''creating attn mask for guided attention'''
+    r'''creating attn mask for guided attention
+    TODO: vectorize'''
     M = np.zeros([N, T])
     for t in range(T):
         for n in range(N):
@@ -141,7 +142,7 @@ def create_attn_mask(N, T, g=0.05):
             M[n, t] = val
     e_x = np.exp(M - np.max(M))
     M = e_x / e_x.sum(axis=0) # only difference
-    M = Variable(torch.FloatTensor(M).t()).cuda()
+    M = torch.FloatTensor(M).t().cuda()
     M = torch.stack([M]*32)
     return M