Re: [theano-users] Error using floatX = float16 to save memory

2016-10-14 Thread luca . wagner . 0812
On Thursday, October 13, 2016 at 5:42:30 PM UTC+2, nouiz wrote: > > For the memory error, the problem is that you try to allocate 14G for a > shared variable on a 12G GPU. This is probably not what you want to do. > > Use theano.tensor.nnet.conv3d now (not conv3d2d.conv3d() or dnn_conv3d). >

Re: [theano-users] Error using floatX = float16 to save memory

2016-10-14 Thread luca . wagner . 0812
Hi Pascal, I don't know how to see what happens during "raise_with_op" in that call. This is the output using pdb inside spyder: Python 2.7.12 |Continuum Analytics, Inc.| (default, Jul 2 2016, 17:42:40) [GCC 4.4.7 20120313 (Red Hat 4.4.7-1)] on linux2 Type "help", "copyright", "credits" or

[theano-users] Theano.clone with Multiple Replacements

2016-10-14 Thread John Coolidge
Hello, I'm trying to use theano.clone to implement dropout in my MLP network. Because I want to apply dropout at multiple layers, I pass the clone call multiple key value pairs to its replacement parameter: replace={layer1:mask*layer1, layer2:mask*layer2, etc} however the graph that's

Re: [theano-users] When I run "import theano" I get "Not able to select available GPU from 2 cards (out of memory)."

2016-10-14 Thread Frédéric Bastien
The order of GPU between nvidia-smi and the cuda driver that Theano use is not always the same. Try with gpu1 as this could the be GPU with memory available. Le 12 oct. 2016 11:34, a écrit : > On this machine I have another process using theano with memory on the > GPU.

Re: [theano-users] Theano.clone with Multiple Replacements

2016-10-14 Thread John Coolidge
I see, thanks Pascal! Shame map_variables doesn't do the trick in this case. I think I'll go with the manual approach you recommended as it seems the most efficient and relatively straight forward in my case. On Friday, October 14, 2016 at 5:13:53 PM UTC-7, Pascal Lamblin wrote: > > On Sat,

Re: [theano-users] When I run "import theano" I get "Not able to select available GPU from 2 cards (out of memory)."

2016-10-14 Thread Frédéric Bastien
Try gpu1 instead of gpu0. The number from nvidia-smi is not always the same completed to cuda driver that Theano use. Le 12 oct. 2016 11:34, a écrit : > On this machine I have another process using theano with memory on the > GPU. Normally I can launch many processes

Re: [theano-users] Theano.clone with Multiple Replacements

2016-10-14 Thread Pascal Lamblin
On Sat, Oct 15, 2016, Pascal Lamblin wrote: > Another option, still experimental, may be the `map_variables` function > in scan_modules/scan_utils. There seem to be some challenges regarding scalar constants with that function, but I was able to do the following: >>>

Re: [theano-users] Theano.clone with Multiple Replacements

2016-10-14 Thread Pascal Lamblin
Hi, Yes, it is an actual problem that we never managed to fix in a satisfactory way. The current behaviour is inconsistent. Doing the substitution one at a time is a workaround, I think Blocks does that for dropout, but it can be cumbersome to have everything cloned over and over again. Another

Re: [theano-users] output of conv layer using stride

2016-10-14 Thread Beatriz G.
What I have done is a bit rudimentary, but maybe it could help to anyone. I have saved the output of the layer where stride (or subsamples, like in the conv2d is called) is produced in a pickle file, and I have load it in a new python file and looking into the output I could know the size and