Mercurial > pylearn

comparison doc/v2_planning/dataset.txt @ 1091:319de699fb67

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
dataset: Reply to GPU question
author Olivier Delalleau <delallea@iro>
date Sun, 12 Sep 2010 14:14:23 -0400
parents f15216356522
children 75175e2e697d
comparison
equal deleted inserted replaced
1090:a80b296eb0df 1091:319de699fb67
303 303
304 RP: I actually like this idea of having the dataset implement the same 304 RP: I actually like this idea of having the dataset implement the same
305 interface as the learner ( or actually a subset of the interface .. ). 305 interface as the learner ( or actually a subset of the interface .. ).
306 I hope people decide to do this. 306 I hope people decide to do this.
307 307
308 308 Support for shared variables
309 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
309 310
310 RP asks: What is the status of having the dataset support copying data 311 RP asks: What is the status of having the dataset support copying data
311 on the GPU ( by storing data in shared variables) ? Have you decided to 312 on the GPU ( by storing data in shared variables) ? Have you decided to
312 include this feature or not ? I think that the strongest selling point of 313 include this feature or not ? I think that the strongest selling point of
313 Theano is that it runs on GPU transperently, and I see this as a good 314 Theano is that it runs on GPU transperently, and I see this as a good
314 selling point for the library as well. Plus we intend to move more and 315 selling point for the library as well. Plus we intend to move more and
315 more towards running things on GPU. If the dataset object does not support 316 more towards running things on GPU. If the dataset object does not support
316 this feature we will need to find hacks around it .. 317 this feature we will need to find hacks around it ..
317 318
319 OD: I have like zero experience with GPU so hopefully someone else can answer
320 this. But the way I see it, hopefully it could work by having some dataset
321 object that would take care of storing its input data into a shared variable.
322