"Duplicate name" warning while dumping a SdA
Closed this issue · 7 comments
There will be User Warning: Duplicate name W and User Warning: Duplicate name b while I dumping my SdA model using theano.misc.pkl_utils.dump. And I found that a SdA model cannot be loaded correctly because of the duplicate name of variables W and b.
So is it necessary to use different name for each theano tensor W and b in model classes? In fact, that's the way I solve the problem mentioned above.
The name is just for the user, so normally, having duplicate name isn't a problem. Can you give me the full stack trace? I'm not able to find it in Theano.
This is the warning.
And I found that theano can't restore SdA model correctly, it will mix up the W and b parameters between the Hidden Layer and Softmax Layer
Update Theano to the development version.. I think it is now fixed.
On Mon, Jun 29, 2015 at 11:25 AM, 王昌旭 notifications@github.com wrote:
This is the warning.
[image: qq20150629-3 2x]
https://cloud.githubusercontent.com/assets/4102875/8411344/1bdf4dd8-1eb6-11e5-9f94-4253da62b720.jpgAnd I found that theano can't restore SdA model correctly, it will mix up
the W and b parameters between the Hidden Layer and Softmax Layer—
Reply to this email directly or view it on GitHub
#95 (comment)
.
OK，thanks.
在 2015年6月29日，23:41，Frédéric Bastien notifications@github.com 写道：
Update Theano to the development version.. I think it is now fixed.
On Mon, Jun 29, 2015 at 11:25 AM, 王昌旭 notifications@github.com wrote:
This is the warning.
[image: qq20150629-3 2x]
https://cloud.githubusercontent.com/assets/4102875/8411344/1bdf4dd8-1eb6-11e5-9f94-4253da62b720.jpgAnd I found that theano can't restore SdA model correctly, it will mix up
the W and b parameters between the Hidden Layer and Softmax Layer―
Reply to this email directly or view it on GitHub
#95 (comment)
.―
Reply to this email directly or view it on GitHub.
BTW, how to upgrade Theano to the development version via pip?
Thanks you.
Frédéric Bastien wrote:
see this page:
http://www.deeplearning.net/software/theano/install.html#bleeding-edge-install-instructions—
Reply to this email directly or view it on GitHub
#95 (comment).