Moved new gpu backend gpuarray from sandbox to main theano directory

doc/library/sandbox/gpuarray/dnn.txt → doc/library/gpuarray/dnn.txt

@@ -135,27 +135,27 @@ To get an error if Theano can not use cuDNN, use this Theano flag:
 Functions
 =========
 
-.. automodule:: theano.sandbox.gpuarray.dnn
+.. automodule:: theano.gpuarray.dnn
    :noindex:
    :members: dnn_conv, dnn_pool
 
 Convolution Ops
 ===============
 
-.. automodule:: theano.sandbox.gpuarray.dnn
+.. automodule:: theano.gpuarray.dnn
    :noindex:
    :members: GpuDnnConvDesc, GpuDnnConv, GpuDnnConvGradW, GpuDnnConvGradI
 
 Pooling Ops
 ===========
 
-.. automodule:: theano.sandbox.gpuarray.dnn
+.. automodule:: theano.gpuarray.dnn
    :noindex:
    :members: GpuDnnPoolDesc, GpuDnnPool, GpuDnnPoolGrad
 
 Softmax Ops
 ===========
 
-.. automodule:: theano.sandbox.gpuarray.dnn
+.. automodule:: theano.gpuarray.dnn
    :noindex:
    :members: GpuDnnSoftmax, GpuDnnSoftmaxGrad

doc/library/sandbox/gpuarray/extra.txt → doc/library/gpuarray/extra.txt

@@ -7,11 +7,11 @@ Utility functions
 Optimisation
 ------------
 
-.. automodule:: theano.sandbox.gpuarray.opt_util
+.. automodule:: theano.gpuarray.opt_util
    :members:
 
 Kernel generation
 -----------------
 
-.. automodule:: theano.sandbox.gpuarray.kernel_codegen
+.. automodule:: theano.gpuarray.kernel_codegen
    :members:

doc/library/sandbox/gpuarray/index.txt → doc/library/gpuarray/index.txt

@@ -5,7 +5,7 @@
 :mod:`gpuarray` -- The (new) GPU backend
 =======================================================
 
-.. module:: theano.sandbox.gpuarray
+.. module:: theano.gpuarray
    :platform: Unix, Windows
    :synopsis: Code for GPU programming (new)
 .. moduleauthor:: MILA

doc/library/sandbox/gpuarray/op.txt → doc/library/gpuarray/op.txt

@@ -13,35 +13,35 @@ is just useful to let people know what is implemented on the gpu.
 Basic Op
 ========
 
-.. automodule:: theano.sandbox.gpuarray.basic_ops
+.. automodule:: theano.gpuarray.basic_ops
     :members:
 
 Blas Op
 =======
 
-.. automodule:: theano.sandbox.gpuarray.blas
+.. automodule:: theano.gpuarray.blas
     :members:
 
-.. automodule:: theano.sandbox.gpuarray.nerv
+.. automodule:: theano.gpuarray.nerv
     :members:
 
 Elemwise Op
 ===========
 
-.. automodule:: theano.sandbox.gpuarray.elemwise
+.. automodule:: theano.gpuarray.elemwise
     :members:
 
 Subtensor Op
 ============
 
-.. automodule:: theano.sandbox.gpuarray.subtensor
+.. automodule:: theano.gpuarray.subtensor
     :members:
 
 Nnet Op
 =======
 
-.. automodule:: theano.sandbox.gpuarray.nnet
+.. automodule:: theano.gpuarray.nnet
     :members:
 
-.. automodule:: theano.sandbox.gpuarray.neighbours
+.. automodule:: theano.gpuarray.neighbours
     :members:

doc/library/sandbox/gpuarray/type.txt → doc/library/gpuarray/type.txt

@@ -4,5 +4,5 @@
 :mod:`gpuarray.type` -- Type classes
 ===================================================
 
-.. automodule:: theano.sandbox.gpuarray.type
+.. automodule:: theano.gpuarray.type
    :members:

doc/library/index.txt

@@ -17,6 +17,7 @@ Types and Ops that you can use to build and compile expression graphs.
    printing
    d3viz/index
    compile/index
+   gpuarray/index
    sparse/index
    sparse/sandbox
    scalar/index

doc/tutorial/using_gpu.txt

@@ -470,7 +470,7 @@ the GPU object directly.  The following code is modifed to do just that.
 
   rng = numpy.random.RandomState(22)
   x = shared(numpy.asarray(rng.rand(vlen), config.floatX))
-  f = function([], sandbox.gpuarray.basic_ops.GpuFromHost(None)(tensor.exp(x)))
+  f = function([], gpuarray.basic_ops.GpuFromHost(None)(tensor.exp(x)))
   print(f.maker.fgraph.toposort())
   t0 = time.time()
   for i in range(iters):
@@ -485,7 +485,7 @@ the GPU object directly.  The following code is modifed to do just that.
   else:
       print('Used the gpu')
 
-Here the :func:`theano.sandbox.gpuarray.basic_ops.GpuFromHost(None)` call
+Here the :func:`theano.gpuarray.basic_ops.GpuFromHost(None)` call
 means "copy input to the GPU", with ``None`` the default GPU context when not
 explicitly given. However during the optimization phase,
 since the result will already be on the gpu, it will be removed.  It is

theano/__init__.py

@@ -116,7 +116,7 @@ if (config.device.startswith('cuda') or
         config.init_gpu_device.startswith('cuda') or
         config.init_gpu_device.startswith('opencl') or
         config.contexts != ''):
-    import theano.sandbox.gpuarray
+    import theano.gpuarray
 
 # Use config.numpy to call numpy.seterr
 import numpy

theano/sandbox/gpuarray/__init__.py → theano/gpuarray/__init__.py

@@ -10,7 +10,7 @@ from theano.compile import optdb
 
 from theano.tensor.basic import register_transfer
 
-_logger_name = 'theano.sandbox.gpuarray'
+_logger_name = 'theano.gpuarray'
 _logger = logging.getLogger(_logger_name)
 
 error = _logger.error

theano/sandbox/gpuarray/blocksparse.py → theano/gpuarray/blocksparse.py

@@ -12,7 +12,7 @@ from theano.gradient import grad_undefined
 from .type import gpu_context_type
 from .basic_ops import as_gpuarray_variable, infer_context_name
 
-_logger = logging.getLogger('theano.sandbox.gpuarray.blocksparse')
+_logger = logging.getLogger('theano.gpuarray.blocksparse')
 
 
 class GpuSparseBlockGemv(COp):

theano/sandbox/gpuarray/comp.py → theano/gpuarray/comp.py

@@ -40,7 +40,7 @@ class NVCC_compiler(NVCC_base):
 
         # If the user didn't specify architecture flags add them
         if not any(['-arch=sm_' in f for f in flags]):
-            dev = theano.sandbox.gpuarray.init_dev.device
+            dev = theano.gpuarray.init_dev.device
             if dev is None:
                 raise Exception("Trying to compile GPU code without a context")
             if dev.startswith("opencl"):

theano/sandbox/gpuarray/multinomial.py → theano/gpuarray/multinomial.py

@@ -12,7 +12,7 @@ import theano.sandbox.multinomial
 from theano import Apply, config
 from theano.gof import Op
 from theano.tensor import NotScalarConstantError, get_scalar_constant_value
-from theano.sandbox import gpuarray
+from theano import gpuarray
 from .basic_ops import as_gpuarray_variable, infer_context_name
 from .opt import register_opt, op_lifter
 from .type import GpuArrayType

theano/sandbox/gpuarray/opt.py → theano/gpuarray/opt.py

@@ -46,7 +46,7 @@ from .subtensor import (GpuIncSubtensor, GpuSubtensor,
                         GpuAdvancedIncSubtensor1_dev20)
 from .opt_util import alpha_merge, output_merge
 
-_logger = logging.getLogger("theano.sandbox.gpuarray.opt")
+_logger = logging.getLogger("theano.gpuarray.opt")
 
 gpu_optimizer = EquilibriumDB()
 gpu_cut_copies = EquilibriumDB()

theano/sandbox/gpuarray/tests/GpuArray.pkl → theano/gpuarray/tests/GpuArray.pkl

-ctheano.sandbox.gpuarray.type
+ctheano.gpuarray.type
 GpuArray_unpickler
 p0
 (cnumpy.core.multiarray

theano/sandbox/gpuarray/tests/config.py → theano/gpuarray/tests/config.py

 from __future__ import absolute_import, print_function, division
 from nose.plugins.skip import SkipTest
 
-import theano.sandbox.gpuarray
+import theano.gpuarray
 
-if theano.sandbox.gpuarray.pygpu is None:
+if theano.gpuarray.pygpu is None:
     raise SkipTest("pygpu not installed")
 
-if (not theano.sandbox.gpuarray.pygpu_activated and
+if (not theano.gpuarray.pygpu_activated and
         not theano.config.init_gpu_device.startswith('gpu')):
-    theano.sandbox.gpuarray.init_dev('cuda')
+    theano.gpuarray.init_dev('cuda')
 
-if not theano.sandbox.gpuarray.pygpu_activated:
+if not theano.gpuarray.pygpu_activated:
     raise SkipTest("pygpu disabled")
 
 test_ctx_name = None

theano/sandbox/gpuarray/tests/test_basic_ops.py → theano/gpuarray/tests/test_basic_ops.py

@@ -302,7 +302,7 @@ class G_reshape(test_basic.T_reshape):
             mode=mode_with_gpu,
             ignore_topo=(HostFromGpu, GpuFromHost,
                          theano.compile.DeepCopyOp,
-                         theano.sandbox.gpuarray.elemwise.GpuElemwise,
+                         theano.gpuarray.elemwise.GpuElemwise,
                          theano.tensor.opt.Shape_i,
                          theano.tensor.opt.MakeVector))
         assert self.op == GpuReshape
@@ -405,7 +405,7 @@ def test_hostfromgpu_shape_i():
                                 'local_dot22_to_dot22scalar',
                                 'specialize')
     a = T.fmatrix('a')
-    ca = theano.sandbox.gpuarray.type.GpuArrayType('float32', (False, False))()
+    ca = theano.gpuarray.type.GpuArrayType('float32', (False, False))()
     av = numpy.asarray(numpy.random.rand(5, 4), dtype='float32')
     cv = gpuarray.asarray(numpy.random.rand(5, 4),
                           dtype='float32',

theano/sandbox/gpuarray/tests/test_opt.py → theano/gpuarray/tests/test_opt.py

@@ -7,7 +7,7 @@ from theano.tests.breakpoint import PdbBreakpoint
 from theano.tests import unittest_tools as utt, test_ifelse
 from theano.tensor.tests import test_basic
 
-import theano.sandbox.gpuarray
+import theano.gpuarray
 from .. import basic_ops
 from ..type import GpuArrayType, gpuarray_shared_constructor, get_context
 from ..basic_ops import (

theano/sandbox/gpuarray/tests/test_pickle.py → theano/gpuarray/tests/test_pickle.py

@@ -14,12 +14,12 @@ from nose.plugins.skip import SkipTest
 from nose.tools import assert_raises
 import numpy
 
-import theano.sandbox.gpuarray
+import theano.gpuarray
 from theano.compat import PY3
 from theano import config
 from theano.misc.pkl_utils import CompatUnpickler
 
-if not theano.sandbox.gpuarray.pygpu_activated:
+if not theano.gpuarray.pygpu_activated:
     try:
         import pygpu
     except ImportError:
@@ -29,7 +29,7 @@ if not theano.sandbox.gpuarray.pygpu_activated:
         cuda_ndarray.use('gpu', default_to_move_computation_to_gpu=False,
                          move_shared_float32_to_gpu=False,
                          enable_cuda=False)
-        theano.sandbox.gpuarray.init_dev('cuda')
+        theano.gpuarray.init_dev('cuda')
 
 from .. import pygpu_activated  # noqa
 

theano/sandbox/gpuarray/type.py → theano/gpuarray/type.py

@@ -233,7 +233,7 @@ class GpuArrayType(Type):
         return data
 
     def filter_variable(self, other, allow_convert=True):
-        from theano.sandbox.gpuarray import GpuFromHost
+        from theano.gpuarray import GpuFromHost
 
         if hasattr(other, '_as_GpuArrayVariable'):
             other = other._as_GpuArrayVariable(self.context_name)

theano/misc/check_multi_gpu.py

@@ -12,8 +12,8 @@ import time
 import numpy
 
 import theano
-from theano.sandbox.gpuarray import init_dev
-from theano.sandbox.gpuarray.blas import gpu_dot22
+from theano.gpuarray import init_dev
+from theano.gpuarray.blas import gpu_dot22
 
 
 def main(dev1, dev2):

theano/misc/may_share_memory.py

@@ -19,7 +19,7 @@ except ImportError:
         return False
 
 from theano.sandbox import cuda
-from theano.sandbox import gpuarray
+from theano import gpuarray
 
 if cuda.cuda_available:
     from theano.sandbox.cuda.type import CudaNdarrayType

theano/sandbox/rng_mrg.py

@@ -24,10 +24,10 @@ from . import multinomial
 
 import theano.sandbox.cuda
 from theano.sandbox.cuda import GpuOp
-from theano.sandbox.gpuarray.basic_ops import GpuKernelBase, Kernel
-from theano.sandbox.gpuarray.type import GpuArrayType
-from theano.sandbox.gpuarray.fp16_help import write_w
-from theano.sandbox.gpuarray.opt import (register_opt as register_gpua,
+from theano.gpuarray.basic_ops import GpuKernelBase, Kernel
+from theano.gpuarray.type import GpuArrayType
+from theano.gpuarray.fp16_help import write_w
+from theano.gpuarray.opt import (register_opt as register_gpua,
                                          host_from_gpu as host_from_gpua)
 if theano.sandbox.cuda.cuda_available:
     from theano.sandbox.cuda import (CudaNdarrayType,

theano/sandbox/tests/test_rng_mrg.py

@@ -366,9 +366,9 @@ def test_consistency_GPUA_serial():
     are the same as the reference (Java) implementation by L'Ecuyer et al.
 
     """
-    from theano.sandbox.gpuarray.tests.test_basic_ops import \
+    from theano.gpuarray.tests.test_basic_ops import \
         mode_with_gpu as mode
-    from theano.sandbox.gpuarray.type import gpuarray_shared_constructor
+    from theano.gpuarray.type import gpuarray_shared_constructor
 
     seed = 12345
     n_samples = 5
@@ -421,9 +421,9 @@ def test_consistency_GPUA_parallel():
     L'Ecuyer et al.
 
     """
-    from theano.sandbox.gpuarray.tests.test_basic_ops import \
+    from theano.gpuarray.tests.test_basic_ops import \
         mode_with_gpu as mode
-    from theano.sandbox.gpuarray.type import gpuarray_shared_constructor
+    from theano.gpuarray.type import gpuarray_shared_constructor
 
     seed = 12345
     n_samples = 5
@@ -1107,9 +1107,9 @@ def test_overflow_gpu_old_backend():
 
 def test_overflow_gpu_new_backend():
     # run with THEANO_FLAGS=mode=FAST_RUN,init_gpu_device=cuda1,device=cpu
-    from theano.sandbox.gpuarray.tests.test_basic_ops import \
+    from theano.gpuarray.tests.test_basic_ops import \
         mode_with_gpu as mode
-    from theano.sandbox.gpuarray.type import gpuarray_shared_constructor
+    from theano.gpuarray.type import gpuarray_shared_constructor
     seed = 12345
     n_substreams = 7
     curr_rstate = numpy.array([seed] * 6, dtype='int32')

theano/scan_module/scan.py

@@ -982,7 +982,8 @@ def scan(fn,
     # the file because that would force on the user some dependencies that we
     # might do not want to. Currently we are working on removing the
     # dependencies on sandbox code completeley.
-    from theano.sandbox import cuda, gpuarray
+    from theano.sandbox import cuda
+    from theano import gpuarray
     if cuda.cuda_available or gpuarray.pygpu_activated:
         # very often we end up in this situation when we want to
         # replace w with w_copy, where w is a GPU variable

theano/scan_module/scan_op.py

@@ -272,7 +272,7 @@ class Scan(PureOp):
         # If scan has the flag 'gpua' set to false (meaning that is shouldn't
         # use the gpuarray gpu backend ), ensure that is has no input and no
         # output with type GpuArrayType
-        from theano.sandbox.gpuarray import GpuArrayType
+        from theano.gpuarray import GpuArrayType
         if not self.info.get("gpua", False):
             for inp in self.inputs:
                 if isinstance(inp.type, GpuArrayType):

theano/scan_module/scan_opt.py

@@ -1008,8 +1008,8 @@ class ScanInplaceOptimizer(Optimizer):
             # gpuarray might be imported but not its GpuAlloc and
             # GpuAllopEmpty ops.
             try:
-                alloc_ops += (theano.sandbox.gpuarray.GpuAlloc,
-                              theano.sandbox.gpuarray.GpuAllocEmpty)
+                alloc_ops += (theano.gpuarray.GpuAlloc,
+                              theano.gpuarray.GpuAllocEmpty)
             except:
                 pass
 

theano/scan_module/scan_utils.py

@@ -151,7 +151,8 @@ def traverse(out, x, x_copy, d, visited=None):
     if out in visited:
         return d
     visited.add(out)
-    from theano.sandbox import cuda, gpuarray
+    from theano.sandbox import cuda
+    from theano import gpuarray
     if out == x:
         if isinstance(x.type, cuda.CudaNdarrayType):
             d[out] = cuda.gpu_from_host(x_copy)

theano/scan_module/tests/test_scan.py

@@ -4939,7 +4939,7 @@ class T_Scan_Gpuarray(unittest.TestCase, ScanGpuTests):
     """
 
     def __init__(self, *args, **kwargs):
-        from theano.sandbox import gpuarray
+        from theano import gpuarray
         self.gpu_backend = gpuarray
 
         # This is unfortunate, but required

theano/tests/test_flake8.py

@@ -89,7 +89,26 @@ whitelist_flake8 = [
     "sandbox/tests/__init__.py",
     "sandbox/cuda/__init__.py",
     "sandbox/cuda/tests/__init__.py",
-    "sandbox/gpuarray/tests/__init__.py",
+    "sandbox/cuda/tests/walltime.py",
+    "sandbox/cuda/tests/test_gradient.py",
+    "sandbox/cuda/tests/test_neighbours.py",
+    "sandbox/cuda/tests/test_conv_cuda_ndarray.py",
+    "sandbox/cuda/tests/test_var.py",
+    "sandbox/cuda/tests/test_opt.py",
+    "sandbox/cuda/tests/test_blas.py",
+    "sandbox/cuda/tests/test_driver.py",
+    "sandbox/cuda/tests/test_rng_curand.py",
+    "sandbox/cuda/tests/test_basic_ops.py",
+    "sandbox/cuda/tests/test_memory.py",
+    "sandbox/cuda/tests/test_mlp.py",
+    "sandbox/cuda/tests/test_bench_loopfusion.py",
+    "sandbox/cuda/tests/test_blocksparse.py",
+    "sandbox/cuda/tests/test_cuda_ndarray.py",
+    "sandbox/cuda/tests/test_tensor_op.py",
+    "sandbox/cuda/tests/test_extra_ops.py",
+    "sandbox/cuda/tests/test_gemmcorr3d.py",
+    "sandbox/cuda/tests/test_viewop.py",
+    "gpuarray/tests/__init__.py",
     "sandbox/scan_module/scan_utils.py",
     "sandbox/scan_module/scan.py",
     "sandbox/scan_module/scan_op.py",
@@ -100,7 +119,7 @@ whitelist_flake8 = [
     "sandbox/linalg/__init__.py",
     "sandbox/linalg/tests/__init__.py",
     "sandbox/linalg/tests/test_linalg.py",
-    "sandbox/gpuarray/__init__.py",
+    "gpuarray/__init__.py",
     "scan_module/scan_utils.py",
     "scan_module/scan_views.py",
     "scan_module/scan.py",