See https://github.com/Theano/Theano/issues/5529 This can't be unrelated. On my system the new backend fails on ALL driver versions (wrong result or crash and Windows won't boot anymore). There is no good driver version.
On Friday, February 10, 2017 at 7:30:59 PM UTC+1, nouiz wrote: > > We seem to hit a bug in some driver version. See this PR: > > https://github.com/Theano/Theano/issues/5530 > > It name the drivers working and not working. I fyour driver version isn't > named there, can you add a comment that name it? > > Fred > > On Wed, Feb 8, 2017 at 3:45 PM Kiuhnm Mnhuik <[email protected] > <javascript:>> wrote: > >> I created an issue on github but still no answer after 21 hours. Can >> someone confirm that this code doesn't work correctly? >> >> import theano >> import theano.tensor as T >> import numpy as np >> >> d = T.vector('d', dtype=floatX) >> d_np = np.array([1, 1, 1, 1, 0, 1, 1, 1, 1, 1], dtype=floatX) >> f = theano.function([d], d.sum()) >> print(f(d_np)) # 9.0 with gpu, but 5.0 with cuda >> >> >> -- >> >> --- >> You received this message because you are subscribed to the Google Groups >> "theano-users" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected] <javascript:>. >> For more options, visit https://groups.google.com/d/optout. >> > -- --- You received this message because you are subscribed to the Google Groups "theano-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
