Merge test_log_softmax and test_log_softmax_opt

theano/sandbox/cuda/tests/test_dnn.py

@@ -437,7 +437,9 @@ def test_pooling_opt():
 
 
 def test_log_softmax():
-    if not cuda.dnn.dnn_available():
+    # This is a test for an optimization that depends on CuDNN v3 or
+    # more recent. Don't test if the CuDNN version is too old.
+    if not cuda.dnn.dnn_available() or cuda.dnn.version() < (3000, 3000):
         raise SkipTest(cuda.dnn.dnn_available.msg)
 
     x = T.ftensor4()
@@ -446,6 +448,12 @@ def test_log_softmax():
 
     f = theano.function([x], log_out, mode=mode_with_gpu)
 
+    # Ensure that the optimization has been applied
+    dnn_softmax_nodes = [n for n in f.maker.fgraph.toposort() if
+                         isinstance(n.op, cuda.dnn.GpuDnnSoftmax)]
+    assert len(dnn_softmax_nodes) == 1
+    assert dnn_softmax_nodes[0].op.algo == "log"
+
     # Ensure that the output of the function is valid
     input_shapes = [(3, 4, 5, 6),
                     (1025, 2, 3, 4),
@@ -467,26 +475,6 @@ def test_log_softmax():
         utt.assert_allclose(out, expected_out)
 
 
-def test_log_softmax_opt():
-    # This is a test for an optimization that depends on CuDNN v3 or
-    # more recent. Don't test if the CuDNN version is too old.
-    if not cuda.dnn.dnn_available() or cuda.dnn.version() < (3000, 3000):
-        raise SkipTest(cuda.dnn.dnn_available.msg)
-
-    x = T.ftensor4()
-    softmax_out = dnn.GpuDnnSoftmax('bc01', 'accurate', 'channel')(x)
-    log_out = T.log(T.as_tensor_variable(softmax_out))
-
-    f = theano.function([x], log_out, mode=mode_with_gpu)
-
-    dnn_softmax_nodes = [n for n in f.maker.fgraph.toposort() if
-                         isinstance(n.op, cuda.dnn.GpuDnnSoftmax)]
-
-    # Ensure that the optimization has been applied
-    assert len(dnn_softmax_nodes) == 1
-    assert dnn_softmax_nodes[0].op.algo == "log"
-
-
 def test_dnn_tag():
     """
     Test that if cudnn isn't avail we crash and that if it is avail, we use it.