Merge pull request #5265 from nouiz/fix_reload_pickle

theano/gof/cmodule.py

@@ -189,7 +189,7 @@ static struct PyModuleDef moduledef = {{
 
     def add_support_code(self, code):
         assert not self.finalized
-        if code not in self.support_code:  # TODO: KLUDGE
+        if code and code not in self.support_code:  # TODO: KLUDGE
             self.support_code.append(code)
 
     def add_function(self, fn):

theano/gof/op.py

@@ -1270,10 +1270,11 @@ class COp(Op):
         if not isinstance(func_files, list):
             func_files = [func_files]
 
-        self.func_files = [self.get_path(f) for f in func_files]
         self.func_name = func_name
-
-        self.load_c_code()
+        # Keep the original name. If we reload old pickle, we want to
+        # find the new path and new version of the file in Theano.
+        self.func_files = func_files
+        self.load_c_code(func_files)
 
         if len(self.code_sections) == 0:
             raise ValueError("No sections where defined in C files")
@@ -1288,12 +1289,13 @@ class COp(Op):
                 raise ValueError('Cannot have an "op_code_cleanup" section '
                                  'and specify the func_name')
 
-    def load_c_code(self):
+    def load_c_code(self, func_files):
         """
         Loads the c code to perform the Op
         """
+        func_files = [self.get_path(f) for f in func_files]
         self.func_codes = []
-        for func_file in self.func_files:
+        for func_file in func_files:
             with open(func_file, 'r') as f:
                 self.func_codes.append(f.read())
 
@@ -1336,7 +1338,7 @@ class COp(Op):
                 if split[0].strip() != '':
                     raise ValueError('Stray code before first #section '
                                      'statement (in file %s): %s' %
-                                     (self.func_files[i], split[0]))
+                                     (func_files[i], split[0]))
 
                 # Separate the code into the proper sections
                 n = 1
@@ -1344,7 +1346,7 @@ class COp(Op):
                     if split[n] not in self.SECTIONS:
                         raise ValueError(
                             "Unknown section type (in file %s): %s" %
-                            (self.func_files[i], split[n]))
+                            (func_files[i], split[n]))
                     if split[n] not in self.code_sections:
                         self.code_sections[split[n]] = ""
                     self.code_sections[split[n]] += split[n + 1]
@@ -1352,7 +1354,7 @@ class COp(Op):
 
             else:
                 raise ValueError("No valid section marker was found in file "
-                                 "%s" % self.func_files[i])
+                                 "%s" % func_files[i])
 
     def get_op_params(self):
         """

theano/gof/type.py

@@ -772,6 +772,15 @@ if (py_%(name)s == NULL) { %(freefunc)s(%(name)s); }
     def __str__(self):
         return "%s{%s}" % (self.__class__.__name__, self.ctype)
 
+    def __setstate__(self, dct):
+        self.__dict__.update(dct)
+        if not hasattr(self, 'headers'):
+            self.headers = ()
+            self.header_dirs = ()
+            self.libraries = ()
+            self.lib_dirs = ()
+            self.extra_support_code = ""
+
 
 class CDataTypeConstant(graph.Constant):
     def merge_signature(self):

theano/sandbox/cuda/dnn.py

@@ -354,26 +354,6 @@ class GpuDnnConv(DnnBase, COp):
         if self.inplace:
             self.destroy_map = {0: [2]}
 
-        # In cuDNN version older than V3, the FFT implementation and the
-        # option to time the different implementations to get the fastest
-        # are both unavailable.
-        if version() < (3000, 3000):
-            if self.algo == 'fft':
-                raise RuntimeError("cuDNN FFT convolution requires cuDNN v3")
-            elif self.algo in ['guess_once', 'guess_on_shape_change']:
-                raise RuntimeError("cuDNN selection of convolution "
-                                   "implementation based on heuristics "
-                                   "requires cuDNN v3")
-            elif self.algo in ['time_once', 'time_on_shape_change']:
-                raise RuntimeError("cuDNN convolution timing requires cuDNN "
-                                   "v3")
-
-        # The fft_tiling implementation is only available from cuDNN V4 onward
-        if version() < (4000, 4000):
-            if self.algo == 'fft_tiling':
-                raise RuntimeError("cuDNN tiled-FFT convolution requires "
-                                   "cuDNN v4 or more recent")
-
         if version() < (5000, 5000):
             if self.algo == 'winograd':
                 raise RuntimeError("cuDNN winograd convolution requires "
@@ -392,6 +372,9 @@ class GpuDnnConv(DnnBase, COp):
                 self.algo = config.dnn.conv.algo_fwd
         if not hasattr(self, 'inplace'):
             self.inplace = False
+        # Work around to reload old pickle.
+        # We need to find the new file name and reload c code.
+        self.load_c_code(["dnn_base.c", "dnn_conv_base.c", "dnn_fwd.c"])
 
     def get_op_params(self):
         if self.inplace:
@@ -656,6 +639,7 @@ class GpuDnnConvGradW(DnnBase, COp):
                 self.algo = config.dnn.conv.algo_bwd_filter
         if not hasattr(self, 'inplace'):
             self.inplace = False
+        self.load_c_code(["dnn_base.c", "dnn_conv_base.c", "dnn_gw.c"])
 
     def grad(self, inp, grads):
         img, top, output, desc, alpha, beta = inp
@@ -865,13 +849,6 @@ class GpuDnnConvGradI(DnnBase, COp):
         if self.inplace:
             self.destroy_map = {0: [2]}
 
-        # The small-workspace implementation is only available from cuDNN V4
-        # onward.
-        if version() < (4000, 4000):
-            if self.algo == 'fft_tiling':
-                raise RuntimeError("cuDNN's tiled-FFT convolution requires "
-                                   "cuDNN v4 or more recent")
-
         if version() < (5000, 5000):
             if self.algo == 'winograd':
                 raise RuntimeError("cuDNN's winograd convolution requires "
@@ -890,6 +867,7 @@ class GpuDnnConvGradI(DnnBase, COp):
                 self.algo = config.dnn.conv.algo_bwd_data
         if not hasattr(self, 'inplace'):
             self.inplace = False
+        self.load_c_code(["dnn_base.c", "dnn_conv_base.c", "dnn_gi.c"])
 
     def grad(self, inp, grads):
         kerns, top, output, desc, alpha, beta = inp
@@ -1465,13 +1443,6 @@ class GpuDnnPoolDesc(GpuOp):
         self.stride = stride
         self.pad = pad
 
-        if self.get_ndim() == 3 and version() < (3000, 3000):
-            raise RuntimeError("cuDNN 3d pooling requires cuDNN v3")
-        if (mode == 'average_exc_pad' and max(pad) > 0 and
-                version() < (4004, 4004)):
-            raise RuntimeError(
-                "cuDNN pooling mode 'average_exc_pad' requires at least v4")
-
     def get_ndim(self):
         return len(self.ws)
 
@@ -2110,9 +2081,6 @@ class GpuDnnSoftmaxBase(DnnBase):
         DnnBase.__init__(self)
         self.tensor_format = tensor_format
 
-        if algo == 'log' and version() < (3000, 3000):
-            raise RuntimeError("cuDNN log-softmax requires cuDNN v3")
-
         assert(algo in ('fast', 'accurate', 'log'))
         self.algo = algo
 
@@ -3179,7 +3147,7 @@ if True:
     @local_optimizer([GpuElemwise, LogSoftmax])
     def local_log_softmax_dnn(node):
         # The log-softmax implementation is only available starting at cuDNN V3
-        if not dnn_available() or version() < (3000, 3000):
+        if not dnn_available():
             return
 
         if (isinstance(node.op, GpuElemwise) and

theano/scalar/basic.py

@@ -3996,6 +3996,7 @@ class Composite(ScalarOp):
     def __getstate__(self):
         rval = dict(self.__dict__)
         rval.pop('_impls', None)
+        rval.pop('prepare_node_called', None)
         del rval['fgraph']
         return rval
 
@@ -4003,6 +4004,7 @@ class Composite(ScalarOp):
         self.__dict__.update(d)
         # We must call init to set fgraph and _impls again, as otherwise
         # self.perform will not work.
+        self.prepare_node_called = set()
         self.init_fgraph()
         self.init_py_impls()
         assert self._c_code