Frédéric Bastien

good idea. On Wed, Aug 17, 2016 at 10:50 AM, Bart van Merriënboer < [email protected]> wrote: > I wouldn't append random letters but just do what e.g. Chrome does when...

Can you give the error message? In Theano doc, there is a section about speed memory thread off. Le 13 avr. 2016 05:35, "PersianNLPer" [email protected] a écrit : > Thank...

What are the motivation to move this outside of blocks-examples? On Mon, Nov 23, 2015 at 10:31 AM, Dzmitry Bahdanau wrote: > > Let us start the procedure of moving...

There is a cuda code go it that too. Just on case this is useful. Some other project keep it on there CPU for parallel computation with the GPU. Le...

With the new GPU back-end, you can pass it as uint8 data. What you describe would work on the current gpu back-end. But you should not "cast" the data, you...

Can we safely and without loss convert those jpeg to a version that libjpeg-turbo support? On Wed, Jan 27, 2016 at 12:09 PM, David Warde-Farley < [email protected]> wrote: > Note...

which file isn't found?

The new back-end have another allocator that is enabled and preallocate 0 by default, but don't release the memory to the driver by default. So the transition should fix that....

> Could you add a few words to the description on which workflow this affects? Is it when dumping DOT files, or when dumping HTML fusion visualization? I amended the...

> All I see is: > > > [XLA] Add the XLA_FLAGS xla_dump_show_fusion_subcomputations. > > It allows to not dump the fusion internal. This allows to display bigger graph. Today...