singa-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From GitBox <...@apache.org>
Subject [GitHub] [incubator-singa] nudles commented on a change in pull request #482: SINGA-474 hardsigmoid operator
Date Wed, 14 Aug 2019 09:19:03 GMT
nudles commented on a change in pull request #482: SINGA-474 hardsigmoid operator
URL: https://github.com/apache/incubator-singa/pull/482#discussion_r313780583
 
 

 ##########
 File path: python/singa/autograd.py
 ##########
 @@ -358,6 +358,51 @@ def relu(x):
     return ReLU()(x)[0]
 
 
+class HardSigmoid(Operation):
+    def __init__(self,alpha=0.2,gamma=0.5):
+        super(HardSigmoid, self).__init__()
+        self.alpha=alpha
+        self.gamma=gamma
+
+    def forward(self, x):
+        """Do forward propgation.
+        #y = max(0, min(1, alpha * x + gamma))
+        Args:
+            x (CTensor): matrix
+        Returns:
+            a CTensor for the result
+        """
+        if training:
+            self.input = x
+
+        x = singa.AddFloat(singa.MultFloat(x,self.alpha),self.gamma)
+        x = singa.ReLU(x)
+        mask = singa.LTFloat(x, 1.0)
+        mask2 = singa.GEFloat(x, 1.0)
+
+        ans = singa.__add__(singa.__mul__(x, mask),mask2)
 
 Review comment:
   relu will turn x into x if x>=0 and 0 otherwise;
   then, you do not need to add 1? 

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services

Mime
View raw message