Merge pull request #812 from nouiz/mixed

.travis.yml

+language: python
+python:
+  - "2.6"
+#  - "2.7"
+#  - "3.2"
+# command to install dependencies
+install: "pip install . --use-mirrors"
+# command to run tests
+script: theano-nose
\ No newline at end of file

doc/tutorial/using_gpu.txt

@@ -44,9 +44,10 @@ file and run it.
     t0 = time.time()
     for i in xrange(iters):
         r = f()
-    print 'Looping %d times took'%iters, time.time() - t0, 'seconds'
+    t1 = time.time()
+    print 'Looping %d times took' % iters, t1 - t0, 'seconds'
     print 'Result is', r
-    print 'Used the','cpu' if numpy.any( [isinstance(x.op,T.Elemwise) for x in f.maker.fgraph.toposort()]) else 'gpu'
+    print 'Used the', 'cpu' if numpy.any([isinstance(x.op, T.Elemwise) for x in f.maker.fgraph.toposort()]) else 'gpu'
 
 The program just computes the exp() of a bunch of random numbers.
 Note that we use the `shared` function to
@@ -100,10 +101,11 @@ after the T.exp(x) is replaced by a GPU version of exp().
     t0 = time.time()
     for i in xrange(iters):
         r = f()
-    print 'Looping %d times took'%iters, time.time() - t0, 'seconds'
+    t1 = time.time()
+    print 'Looping %d times took' % iters, t1 - t0, 'seconds'
     print 'Result is', r
     print 'Numpy result is', numpy.asarray(r)
-    print 'Used the','cpu' if numpy.any( [isinstance(x.op,T.Elemwise) for x in f.maker.fgraph.toposort()]) else 'gpu'
+    print 'Used the', 'cpu' if numpy.any([isinstance(x.op, T.Elemwise) for x in f.maker.fgraph.toposort()]) else 'gpu'
 
 The output from this program is
 
@@ -155,10 +157,11 @@ that it has the un-wanted side-effect of really slowing things down.
     t0 = time.time()
     for i in xrange(iters):
         r = f()
-    print 'Looping %d times took'%iters, time.time() - t0, 'seconds'
+    t1 = time.time()
+    print 'Looping %d times took' % iters, t1 - t0, 'seconds'
     print 'Result is', r
     print 'Numpy result is', numpy.asarray(r)
-    print 'Used the','cpu' if numpy.any( [isinstance(x.op,T.Elemwise) for x in f.maker.fgraph.toposort()]) else 'gpu'
+    print 'Used the', 'cpu' if numpy.any([isinstance(x.op,T.Elemwise) for x in f.maker.fgraph.toposort()]) else 'gpu'
 
 Running this version of the code takes just under 0.05 seconds, over 140x faster than
 the CPU implementation!

theano/configdefaults_late.py

@@ -7,6 +7,7 @@ This is needed as we need to have parsed the previous
 import os
 import logging
 import subprocess
+import tempfile
 
 import theano
 from theano.configparser import (
@@ -63,16 +64,19 @@ int main( int argc, const char* argv[] )
         }
 }
         """
-        p = os.path.join(config.compiledir, 'test_omp.c')
-        f = open(p, 'w')
-        f.write(code)
-        f.close()
-        p = subprocess.Popen(['g++', '-fopenmp', p], stdout=subprocess.PIPE,
-                         stderr=subprocess.PIPE,
-                         stdin=dummy_stdin.fileno())
-        p.wait()
-        if p.returncode != 0:
-            default_openmp = False
+        fd, path = tempfile.mkstemp(suffix='.c', prefix='test_omp_')
+        try:
+            os.write(fd, code)
+            os.close(fd)
+            proc = subprocess.Popen(['g++', '-fopenmp', path],
+                                    stdout=subprocess.PIPE,
+                                    stderr=subprocess.PIPE,
+                                    stdin=dummy_stdin.fileno())
+            proc.wait()
+            if proc.returncode != 0:
+                default_openmp = False
+        finally:
+            os.remove(path)
     except OSError, e:
         default_openmp = False
 

theano/misc/do_nightly_build_send

@@ -14,7 +14,7 @@ files=["/tmp/do_nightly_build_theano", "/tmp/do_nightly_build_pylearn",
        "/tmp/do_nightly_build_deeplearning", "/tmp/do_nightly_build_pylearn2",
        "/tmp/do_nightly_build_theano_python2.4"]
 msgs=['Theano buildbot', 'Pylearn buildbot', 'Deep Learning Tutorial buildbot',
-      'Pylearn2 buildbot', 'Theano Python2.4']
+      'Pylearn2 buildbot', 'Theano Python2.4 buildbot']
 
 print files
 print msgs

theano/sandbox/cuda/basic_ops.py

@@ -79,6 +79,19 @@ class HostFromGpu(GpuOp):
 
     def infer_shape(self, node, xshp):
         return xshp
+
+    def c_code(self, node, name, inputs, outputs, sub):
+        inp = inputs[0]
+        out = outputs[0]
+        fail = sub['fail']
+        return """
+        %(out)s = (PyArrayObject *) CudaNdarray_CreateArrayObj(%(inp)s);
+        if(!%(out)s)
+            %(fail)s;
+        """ % locals()
+
+    def c_code_cache_version(self):
+        return (1,)
 host_from_gpu = HostFromGpu()
 
 

theano/sandbox/cuda/nvcc_compiler.py

@@ -300,10 +300,6 @@ class NVCC_compiler(object):
         finally:
             os.chdir(orig_dir)
 
-        if nvcc_stdout:
-            # this doesn't happen to my knowledge
-            print >> sys.stderr, "DEBUG: nvcc STDOUT", nvcc_stdout
-
         for eline in nvcc_stderr.split('\n'):
             if not eline:
                 continue
@@ -334,11 +330,16 @@ class NVCC_compiler(object):
                 except Exception:
                     pass
                 print >> sys.stderr, l
+            print nvcc_stdout
             raise Exception('nvcc return status', p.returncode,
                             'for cmd', ' '.join(cmd))
         elif config.cmodule.compilation_warning and nvcc_stdout:
             print nvcc_stdout
 
+        if nvcc_stdout:
+            # this doesn't happen to my knowledge
+            print >> sys.stderr, "DEBUG: nvcc STDOUT", nvcc_stdout
+
         #touch the __init__ file
         file(os.path.join(location, "__init__.py"), 'w').close()
         return dlimport(lib_filename)

theano/tensor/basic.py

@@ -2183,6 +2183,11 @@ class MaxAndArgmax(Op):
 
     def make_node(self, x, axis=None):
         x = _as_tensor_variable(x)
+        if isinstance(axis, Variable):
+            if not isinstance(axis, Constant):
+                raise TypeError("MaxAndArgmax need a constant axis")
+            axis = [axis.data]
+
         if isinstance(axis, int):
             axis = [axis]
         elif isinstance(axis, (tuple, list)):
@@ -2192,6 +2197,7 @@ class MaxAndArgmax(Op):
                 assert axis == range(x.type.ndim), (
                     "MaxAndArgmax does not support multiple"
                     " axes. the max fct supports it.")
+
         # we make the axis all positive to make the infer_shape work
         # with negative axis
         if x.type.ndim > 0 and axis is not None:

theano/tensor/tests/test_basic.py

@@ -1790,7 +1790,8 @@ class T_max_and_argmax(unittest.TestCase):
         data = rand(2, 3)
         n = as_tensor_variable(data)
         for (axis, np_axis)  in [(-1, -1), (0, 0), (1, 1), (None, None),
-                                 ([0, 1], None), ([1, 0], None)]:
+                                 ([0, 1], None), ([1, 0], None),
+                                 (constant(0), 0)]:
             v, i = eval_outputs(max_and_argmax(n, axis))
             assert i.dtype == 'int64'
             self.assertTrue(numpy.all(v == numpy.max(data, np_axis)))

theano/tests/test_tutorial.py

@@ -800,16 +800,17 @@ class T_using_gpu(unittest.TestCase):
         t0 = time.time()
         for i in xrange(iters):
             r = f()
-        print 'Looping %d times took'%iters, time.time() - t0, 'seconds'
+        t1 = time.time()
+        print 'Looping %d times took' % iters, t1 - t0, 'seconds'
         print 'Result is', r
-        if numpy.any( [isinstance(x.op,T.Elemwise) for x in f.maker.fgraph.toposort()]):
+        if numpy.any([isinstance(x.op, T.Elemwise) for x in f.maker.fgraph.toposort()]):
             print 'Used the cpu'
         else:
             print 'Used the gpu'
         if theano.config.device.find('gpu') > -1:
             assert not numpy.any( [isinstance(x.op,T.Elemwise) for x in f.maker.fgraph.toposort()])
         else:
-            assert numpy.any( [isinstance(x.op,T.Elemwise) for x in f.maker.fgraph.toposort()])
+            assert numpy.any([isinstance(x.op, T.Elemwise) for x in f.maker.fgraph.toposort()])
 
 
 
@@ -831,15 +832,16 @@ class T_using_gpu(unittest.TestCase):
             t0 = time.time()
             for i in xrange(iters):
                 r = f()
-            print 'Looping %d times took'%iters, time.time() - t0, 'seconds'
+            t1 = time.time()
+            print 'Looping %d times took' % iters, t1 - t0, 'seconds'
             print 'Result is', r
             print 'Numpy result is', numpy.asarray(r)
-            if numpy.any( [isinstance(x.op,T.Elemwise) for x in f.maker.fgraph.toposort()]):
+            if numpy.any([isinstance(x.op, T.Elemwise) for x in f.maker.fgraph.toposort()]):
                 print 'Used the cpu'
             else:
                 print 'Used the gpu'
 
-            assert not numpy.any( [isinstance(x.op,T.Elemwise) for x in f.maker.fgraph.toposort()])
+            assert not numpy.any([isinstance(x.op, T.Elemwise) for x in f.maker.fgraph.toposort()])
 
 
 
@@ -865,15 +867,16 @@ class T_using_gpu(unittest.TestCase):
             t0 = time.time()
             for i in xrange(iters):
                 r = f()
-            print 'Looping %d times took'%iters, time.time() - t0, 'seconds'
+            t1 = time.time()
+            print 'Looping %d times took' % iters, t1 - t0, 'seconds'
             print 'Result is', r
             print 'Numpy result is', numpy.asarray(r)
-            if numpy.any( [isinstance(x.op,T.Elemwise) for x in f.maker.fgraph.toposort()]):
+            if numpy.any([isinstance(x.op, T.Elemwise) for x in f.maker.fgraph.toposort()]):
                 print 'Used the cpu'
             else:
                 print 'Used the gpu'
 
-            assert not numpy.any( [isinstance(x.op,T.Elemwise) for x in f.maker.fgraph.toposort()])
+            assert not numpy.any([isinstance(x.op, T.Elemwise) for x in f.maker.fgraph.toposort()])
 
 
 class T_fibby(unittest.TestCase):