New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Bug: gradient of convolution in newest theano from master #3763
Comments
I ran |
I think you can use theano.tensor.nnet.conv.conv2d until we fix this. thanks for the report. On Thu, Dec 10, 2015 at 2:02 PM, matthias-k notifications@github.com
|
Just to let you know that we merged the fix. thanks for the report. |
Great, thanks for the quick bugfix! |
I found a strange problem with the newest theano from master. I implemented a gaussian convolution. In theano 0.7.0, taking the gradient of the output with respect to the kernel size works perfectly, however with the newest master (0.7.0.dev-e521b20e578c033d51e548181bd1edd24af64427) I get an execption
ValueError: ('You cannot drop a non-broadcastable dimension.', ((True, False, False, False), (2, 3)))
Here is a minimal example to reproduce the problem:
It fails (with python2.7 and python3.4) in the last line as follows:
The text was updated successfully, but these errors were encountered: