Merge pull request #4726 from abergeron/fix_buildbot

theano/gpuarray/tests/test_pickle.py

@@ -18,6 +18,8 @@ from theano.compat import PY3
 from theano import config
 from theano.misc.pkl_utils import CompatUnpickler
 
+from ..type import ContextNotDefined
+
 try:
     from . import config as _  # noqa
     have_pygpu = True
@@ -26,11 +28,9 @@ except SkipTest:
 
 
 def test_unpickle_gpuarray_as_numpy_ndarray_flag1():
-    """Only test when pygpu isn't
-    available. test_unpickle_gpuarray_as_numpy_ndarray_flag0 in
-    test_type.py test it when pygpu is there.
-
-    """
+    # Only test when pygpu isn't
+    # available. test_unpickle_gpuarray_as_numpy_ndarray_flag0 in
+    # test_type.py test it when pygpu is there.
     if have_pygpu:
         raise SkipTest("pygpu active")
     oldflag = config.experimental.unpickle_gpu_on_cpu
@@ -45,7 +45,7 @@ def test_unpickle_gpuarray_as_numpy_ndarray_flag1():
                 u = CompatUnpickler(fp, encoding="latin1")
             else:
                 u = CompatUnpickler(fp)
-            assert_raises(ImportError, u.load)
+            assert_raises((ImportError, ContextNotDefined), u.load)
     finally:
         config.experimental.unpickle_gpu_on_cpu = oldflag
 

theano/sandbox/cuda/tests/test_dnn.py

@@ -723,7 +723,7 @@ def test_batchnorm_train():
     utt.seed_rng()
 
     for mode in ('per-activation', 'spatial'):
-        for vartype in (T.tensor4, T.tensor3, T.matrix, T.vector):
+        for vartype in (T.ftensor4, T.ftensor3, T.fmatrix, T.fvector):
             x, scale, bias = (vartype(n) for n in ('x', 'scale', 'bias'))
             ndim = x.ndim
             eps = 5e-3  # some non-standard value to test if it's used
@@ -751,15 +751,14 @@ def test_batchnorm_train():
                                 [out, x_mean, x_invstd, out2, x_mean2, x_invstd2] +
                                 grads + grads2, mode=mode_with_gpu)
             # run
-            floatX = theano.config.floatX
             for data_shape in ((10, 20, 30, 40), (4, 3, 1, 1), (1, 1, 5, 5)):
                 data_shape = data_shape[:ndim]
                 param_shape = tuple(1 if d in axes else s
                                     for d, s in enumerate(data_shape))
-                X = 4 + 3 * numpy.random.randn(*data_shape).astype(floatX)
-                Dy = -1 + 2 * numpy.random.randn(*data_shape).astype(floatX)
-                Scale = numpy.random.randn(*param_shape).astype(floatX)
-                Bias = numpy.random.randn(*param_shape).astype(floatX)
+                X = 4 + 3 * numpy.random.randn(*data_shape).astype('float32')
+                Dy = -1 + 2 * numpy.random.randn(*data_shape).astype('float32')
+                Scale = numpy.random.randn(*param_shape).astype('float32')
+                Bias = numpy.random.randn(*param_shape).astype('float32')
                 outputs = f(X, Scale, Bias, Dy)
                 # compare outputs
                 utt.assert_allclose(outputs[0], outputs[0 + 3])  # out
@@ -779,7 +778,7 @@ def test_batchnorm_inference():
     utt.seed_rng()
 
     for mode in ('per-activation', 'spatial'):
-        for vartype in (T.tensor4, T.tensor3, T.matrix, T.vector):
+        for vartype in (T.ftensor4, T.ftensor3, T.fmatrix, T.fvector):
             x, scale, bias, mean, var = (vartype(n) for n in ('x', 'scale',
                                                               'bias', 'mean',
                                                               'var'))
@@ -806,17 +805,16 @@ def test_batchnorm_inference():
             f = theano.function([x, scale, bias, mean, var, dy],
                                 [out, out2] + grads + grads2, mode=mode_with_gpu)
             # run
-            floatX = theano.config.floatX
             for data_shape in ((10, 20, 30, 40), (4, 3, 1, 1), (1, 1, 5, 5)):
                 data_shape = data_shape[:ndim]
                 param_shape = tuple(1 if d in axes else s
                                     for d, s in enumerate(data_shape))
-                X = 4 + 3 * numpy.random.randn(*data_shape).astype(floatX)
-                Dy = -1 + 2 * numpy.random.randn(*data_shape).astype(floatX)
-                Scale = numpy.random.randn(*param_shape).astype(floatX)
-                Bias = numpy.random.randn(*param_shape).astype(floatX)
-                Mean = numpy.random.randn(*param_shape).astype(floatX)
-                Var = numpy.random.rand(*param_shape).astype(floatX)
+                X = 4 + 3 * numpy.random.randn(*data_shape).astype('float32')
+                Dy = -1 + 2 * numpy.random.randn(*data_shape).astype('float32')
+                Scale = numpy.random.randn(*param_shape).astype('float32')
+                Bias = numpy.random.randn(*param_shape).astype('float32')
+                Mean = numpy.random.randn(*param_shape).astype('float32')
+                Var = numpy.random.rand(*param_shape).astype('float32')
                 outputs = f(X, Scale, Bias, Mean, Var, Dy)
                 # compare outputs
                 utt.assert_allclose(outputs[0], outputs[1])  # out

theano/tensor/signal/tests/test_pool.py

@@ -35,8 +35,8 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
             if input.shape[-1] % ds[1]:
                 yi += 1
         out_shp = list(input.shape[:-2])
-        out_shp.append(input.shape[-2] / ds[0] + xi)
-        out_shp.append(input.shape[-1] / ds[1] + yi)
+        out_shp.append(input.shape[-2] // ds[0] + xi)
+        out_shp.append(input.shape[-1] // ds[1] + yi)
         output_val = numpy.zeros(out_shp)
         func = numpy.max
         if mode == 'sum':