Add Concat

Author: lantiga

test/base.py

@@ -38,15 +38,20 @@ def base_test():
     model7 = lambda x: fc3(F.max_pool2d(x.unsqueeze(dim=0),2).squeeze(dim=0))
     model8 = lambda x: fc3(F.max_pool3d(x.unsqueeze(0),2).squeeze())
     model9 = lambda x: fc3(F.max_pool1d(x.abs().view(1,1,-1),4).squeeze().view(10,10))
+    #model10 = lambda x: fc3(x.double())
+    #model10 = lambda x: fc3(x.view(1,10,10).select(0,0))
+    model10 = lambda x, y: F.softmax(F.tanh(fc3(torch.cat((x,y),1))))
 
     data = Variable(torch.rand(10,10))
     data2 = Variable(torch.rand(20,20))
+    data1a = Variable(torch.rand(10,5))
+    data1b = Variable(torch.rand(10,5))
     data3 = Variable(torch.rand(2,20,20))
 
     out = model0(data) + \
           model1(data) * model2(data) / model3(data) / 2.0 + \
           2.0 * model4(data) + model5(data) + 1 - 2.0 + \
-          model6(data2) + model7(data2) + model8(data3) + model9(data2)
+          model6(data2) + model7(data2) + model8(data3) + model9(data2) + model10(data1a,data1b)
 
     out_path = 'out'
     if not os.path.isdir(out_path):

torch2c/emitters.py

@@ -685,6 +685,36 @@ def free_tpl(self):
 register(Clone, torch.autograd._functions.tensor.Clone)
 
 
+class Concat(Emitter):
+
+    def __init__(self, obj, prevfns):
+        Emitter.__init__(self, obj, prevfns)
+        for i, el in enumerate(prevfns):
+            self.def_vars({'input%d' % i: id(el)})
+        self.ninputs = len(prevfns)
+        self.infer_type_var = 'input0'
+        self.def_args({
+            'dim': obj.dim,
+            'ninputs': self.ninputs
+        })
+
+    def call_tpl(self):
+        arraytpl = '\n'.join(['inputs[%d] = $input%d;' % (i,i) for i in range(self.ninputs)])
+        return '''
+            TH${T}Tensor *$id = TH${T}Tensor_new();
+            TH${T}Tensor *inputs[%d];
+            %s
+            TH${T}Tensor_catArray($id,inputs,$ninputs,$dim);
+            ''' % (self.ninputs, arraytpl)
+
+    def free_tpl(self):
+        return '''
+            TH${T}Tensor_free($id);
+            '''
+
+register(Concat, torch.autograd._functions.tensor.Concat)
+
+
 class Sigmoid(Emitter):
 
     def __init__(self, obj, prevfns):