Use _allclose for comparison, since it is more lax on float32 data.

theano/sandbox/cuda/tests/test_cuda_ndarray.py

@@ -230,7 +230,7 @@ def test_dot():
     b0 = cuda_ndarray.CudaNdarray(a0)
     b1 = cuda_ndarray.CudaNdarray(a1)
 
-    assert numpy.allclose(numpy.dot(a0, a1), cuda_ndarray.dot(b0, b1))
+    assert _allclose(numpy.dot(a0, a1), cuda_ndarray.dot(b0, b1))
 
 
     a1 = theano._asarray(rng.randn(6, 7), dtype='float32')
@@ -240,7 +240,7 @@ def test_dot():
     transposed = cuda_ndarray.dimshuffle(b1,(1,0))
     cuda_version  =  cuda_ndarray.dot(b0,  transposed)
 
-    assert numpy.allclose( numpy_version,  cuda_version)
+    assert _allclose(numpy_version, cuda_version)
 
     a1 = theano._asarray(rng.randn(7, 6), dtype='float32')
     b1 = cuda_ndarray.CudaNdarray(a1)
@@ -249,12 +249,15 @@ def test_dot():
     a0 = theano._asarray(rng.randn(7, 4), dtype='float32')
     b0 = cuda_ndarray.CudaNdarray(a0)
 
-    assert numpy.allclose(numpy.dot(a0.T, a1), cuda_ndarray.dot( cuda_ndarray.dimshuffle(b0,(1,0)), b1))
+    assert _allclose(numpy.dot(a0.T, a1),
+            cuda_ndarray.dot(cuda_ndarray.dimshuffle(b0,(1,0)), b1))
 
     a1 = theano._asarray(rng.randn(6, 7), dtype='float32')
     b1 = cuda_ndarray.CudaNdarray(a1)
 
-    assert numpy.allclose(numpy.dot(a0.T, a1.T), cuda_ndarray.dot( cuda_ndarray.dimshuffle(b0,(1,0)), cuda_ndarray.dimshuffle(b1,(1,0) ) ) )
+    assert _allclose(numpy.dot(a0.T, a1.T),
+            cuda_ndarray.dot(cuda_ndarray.dimshuffle(b0,(1,0)),
+                             cuda_ndarray.dimshuffle(b1,(1,0))))
 
 
 def test_sum():