decreased the sizes of the matrices to make debug mode run faster on those

theano/tests/test_tutorial.py

@@ -763,8 +763,8 @@ class T_using_gpu(unittest.TestCase):
         import numpy
         import time
 
-        vlen = 10 * 30 * 768  # 10 x #cores x # threads per core
-        iters = 1000
+        vlen = 10 * 30 * 70  # 10 x #cores x # threads per core
+        iters = 10
 
         rng = numpy.random.RandomState(22)
         x = shared(numpy.asarray(rng.rand(vlen), config.floatX))
@@ -791,8 +791,8 @@ class T_using_gpu(unittest.TestCase):
             import numpy
             import time
 
-            vlen = 10 * 30 * 768  # 10 x #cores x # threads per core
-            iters = 1000
+            vlen = 10 * 30 * 70  # 10 x #cores x # threads per core
+            iters = 10
 
             rng = numpy.random.RandomState(22)
             x = shared(numpy.asarray(rng.rand(vlen), config.floatX))
@@ -820,8 +820,8 @@ class T_using_gpu(unittest.TestCase):
             import numpy
             import time
 
-            vlen = 10 * 30 * 768  # 10 x #cores x # threads per core
-            iters = 1000
+            vlen = 10 * 30 * 70  # 10 x #cores x # threads per core
+            iters = 10
 
             rng = numpy.random.RandomState(22)
             x = shared(numpy.asarray(rng.rand(vlen), config.floatX))