[GitHub] [incubator-singa] nudles commented on a change in pull request #475: SINGA-463 fixed python test_operation

[GitBox <gi...@apache.org>]

nudles commented on a change in pull request #475: SINGA-463 fixed python test_operation
URL: https://github.com/apache/incubator-singa/pull/475#discussion_r299836848
 
 

 ##########
 File path: python/singa/autograd.py
 ##########
 @@ -924,17 +924,14 @@ def __init__(
         )
         self.W.gaussian(0.0, std)
 
-        if self.bias:
-            b_shape = (self.out_channels,)
-            self.b = Tensor(shape=b_shape, requires_grad=True, stores_grad=True)
-            self.b.set_value(0.0)
-        else:
-            # to keep consistency when to do forward.
-            self.b = None
-            # Tensor(data=CTensor([]), requires_grad=False, stores_grad=False)
+        b_shape = (self.out_channels,)
+        # forward method required bias variable to proceed (dummy b tensor)
+        self.b = Tensor(shape=b_shape, requires_grad=self.bias, stores_grad=self.bias)
+        self.b.set_value(0.0)
 
 Review comment:
   the bias tensor cannot be created here because onnx will consider it a real bias (onnx has no bias_term flag).
   [this line](https://github.com/apache/incubator-singa/blob/b99aaa9f87ccbd4cf111e76013b8387b7129bd8b/python/singa/autograd.py#L791) should also be updated to set b as an optional argument `forward(self, x, W, b=None)`.
   You can create the tensor for b before calling the CPP function.

----------------------------------------------------------------
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