Add pygpu and CKernelBase headers in GPU CTC implementation

theano/gpuarray/ctc.py

@@ -12,6 +12,9 @@ from .type import GpuArrayType
 from .opt import register_opt, op_lifter, register_opt2
 from theano.gradient import grad_undefined
 
+import os
+import pygpu
+
 ctc_enabled = config.ctc.enabled
 
 
@@ -30,6 +33,7 @@ class GpuConnectionistTemporalClassification(CGpuKernelBase, Op):
         self.compute_grad = compute_grad
         self.context_name = context_name
 
+        Op.__init__(self)
         CGpuKernelBase.__init__(self, self.func_file, self.func_name)
 
         self.costs_type = GpuArrayType(dtype='float32',
@@ -63,15 +67,20 @@ class GpuConnectionistTemporalClassification(CGpuKernelBase, Op):
             # We assume here that the header is available at the include directory
             # of the CTC root directory.
             dirs.append(os.path.join(config.ctc.root, "include"))
-        return dirs + CGpuKernelBase.c_header_dirs(self)
+        dirs = dirs + list(pygpu.get_include())
+        dirs = dirs + list(super(CGpuKernelBase, self).c_header_dirs())
+        return dirs
 
     def c_headers(self):
-        return ["ctc.h"] + CGpuKernelBase.c_headers(self)
+        headers = ['ctc.h']
+        headers = headers + super(CGpuKernelBase, self).c_headers()
+        headers = headers + ['<numpy_compat.h>', '<gpuarray_helper.h>']
+        return headers
 
     def make_node(self, activations, labels, input_lengths):
         if not ctc_enabled:
             raise RuntimeError('Baidu CTC is not enabled and '
-                               'ConnectionistTemporalClassification Op '
+                               'GpuConnectionistTemporalClassification Op '
                                'can not be constructed.')
 
         context = infer_context_name(activations, labels, input_lengths)
@@ -106,10 +115,10 @@ class GpuConnectionistTemporalClassification(CGpuKernelBase, Op):
                             outputs=out_params)
 
     def grad(self, inputs, output_grads):
-        return [grad_undefined(self, 0, inputs[0]),
+        return [self.grads_type(),
                 grad_undefined(self, 1, inputs[1]),
                 grad_undefined(self, 2, inputs[2])]
 
 def ctc(activations, labels, input_lengths):
     return GpuConnectionistTemporalClassification()(activations, labels,
-                                                    input_lengths)
\ No newline at end of file
+                                                    input_lengths)