add conv3d2d optimizer

theano/gpuarray/opt.py

@@ -24,7 +24,7 @@ from theano.scalar.basic import log, neg, true_div
 from theano.scalar.basic_scipy import Erfinv, Erfcinv
 from theano.scan_module import scan_utils, scan_op, scan_opt
 
-from theano.tensor.nnet import bn
+from theano.tensor.nnet import bn, conv3d2d
 from theano.tensor.nnet.conv import ConvOp
 from theano.tensor.nnet.blocksparse import SparseBlockGemv, SparseBlockOuter
 from theano.tensor.nnet.abstract_conv import (BaseAbstractConv,
@@ -1739,6 +1739,32 @@ def local_abstractconv3d_gemm(node):
     return [rval]
 
 
+@local_optimizer([AbstractConv3d])
+def local_abstractconv3d2d(node):
+    if not isinstance(node.op, AbstractConv3d):
+        return None
+    img, kern = node.inputs
+    if (not isinstance(img.type, GpuArrayType) or
+            not isinstance(kern.type, GpuArrayType)):
+        return None
+
+    ctx = infer_context_name(img, kern)
+    border_mode = node.op.border_mode
+    subsample = node.op.subsample
+    filter_dilation = node.op.filter_dilation
+
+    if subsample == (1, 1, 1) and filter_dilation == (1, 1, 1):
+        rval = conv3d2d.conv3d(gpu_contiguous(img.dimshuffle(0, 2, 1, 3, 4)),
+                               gpu_contiguous(kern.dimshuffle(0, 2, 1, 3, 4)),
+                               border_mode=border_mode)
+        rval = as_gpuarray_variable(rval.dimshuffle(0, 2, 1, 3, 4),
+                                    context_name=ctx)
+
+        return [rval]
+    else:
+        return None
+
+
 @local_optimizer([AbstractConv2d_gradWeights])
 def local_abstractconv_gradweights_gemm(node):
     if not isinstance(node.op, AbstractConv2d_gradWeights):
@@ -2578,6 +2604,7 @@ conv_metaopt.register([local_abstractconv_gemm_alternative])
 conv_metaopt.register([local_abstractconv_gemm_gradweights_alt])
 conv_metaopt.register([local_abstractconv_gradinputs_gemm_alt])
 conv_metaopt.register([local_abstractconv_cudnn_alternative])
+conv_metaopt.register([local_abstractconv3d2d])
 abstractconv_groupopt.register('conv_metaopt', conv_metaopt, 'conv_meta', position=0)
 
 # Register cuDNN batch normalization implementation