Klaus Greff
Klaus Greff
What would be the goal of this? It seems like a hack that only works in certain cases. I think if we tackle the issue of retaining context we should...
How about putting the network in a special keep-context mode, and then having a hook call clear context on it when needed? That would work for training but not so...
Correct. At some point we've used `codes` to see how well we could classify textures in the image. It is also not used for training. So just using arrays of...
We haven't seen a case of overfitting yet. However for some reason the shapes network needs to be rather big. You can run with fewer units, but performance does degrade.
Hi, sounds like a theano problem with float64. Do you have a `.theanorc` file in your home with this? ``` [global] warn_float64 = raise floatX = float32 ```
The failure is this, and it can randomly occurr both in python 2 and 3. It seems to be much more frequent on the test server than on my local...
The problem seems to persist, so I am reopening. :(
Wouldn't this explanation lead to an infinite recursion with duplicated output? The strange thing about this is, that I cannot reproduce this issue at all on my machine. For me...
@yulonglin I have not looked into this any further, as I am currently unable to use sacred in my pipeline. If possible I'd recommend to use `capture_mode="sys"` which appears to...
This is troubling, and I currently don't have any idea what could cause this behavior. If you find any further hints, please let me know.