removed unused functions

test/test_spline_conv.py

@@ -31,15 +31,14 @@ def test_spline_conv_cpu(tensor):
     ]
 
     expected_output = [
-        (1 + 12.5 * 9 + 13 * 10 + sum(edgewise_output)) / 5,
-        1 + 12.5 * 1 + 13 * 2,
-        1 + 12.5 * 3 + 13 * 4,
-        1 + 12.5 * 5 + 13 * 6,
-        1 + 12.5 * 7 + 13 * 8,
+        [1 + 12.5 * 9 + 13 * 10 + sum(edgewise_output) / 4],
+        [1 + 12.5 * 1 + 13 * 2],
+        [1 + 12.5 * 3 + 13 * 4],
+        [1 + 12.5 * 5 + 13 * 6],
+        [1 + 12.5 * 7 + 13 * 8],
     ]
 
-    output = [pytest.approx(o, 0.01) for o in output.view(-1).tolist()]
-    assert output == expected_output
+    assert output.tolist() == expected_output
 
     x, weight, pseudo = Variable(x), Variable(weight), Variable(pseudo)
     root_weight, bias = Variable(root_weight), Variable(bias)
@@ -47,8 +46,7 @@ def test_spline_conv_cpu(tensor):
     output = spline_conv(x, edge_index, pseudo, weight, kernel_size,
                          is_open_spline, 1, root_weight, bias)
 
-    output = [pytest.approx(o, 0.01) for o in output.data.view(-1).tolist()]
-    assert output == expected_output
+    assert output.data.tolist() == expected_output
 
 
 def test_spline_weighting_backward_cpu():

torch_spline_conv/functions/spline_conv.py

@@ -15,39 +15,12 @@ def spline_conv(x,
                 root_weight=None,
                 bias=None):
 
-    # Convolve over each node.
-    output = basic_spline_conv(x, edge_index, pseudo, weight, kernel_size,
-                               is_open_spline, degree)
-
-    # Compute degree.
-    degree = x.new() if torch.is_tensor(x) else x.data.new()
-    degree = node_degree(edge_index, x.size(0), out=degree)
-
-    # Weight root node separately (if wished).
-    if root_weight is not None:
-        output += torch.mm(x, root_weight)
-        degree += 1
-
-    # Normalize output by node degree.
-    degree = degree.unsqueeze(-1).clamp_(min=1)
-    output /= degree if torch.is_tensor(x) else Var(degree)
-
-    # Add bias (if wished).
-    if bias is not None:
-        output += bias
-
-    return output
-
-
-def basic_spline_conv(x, edge_index, pseudo, weight, kernel_size,
-                      is_open_spline, degree):
-
     n, e, m_out = x.size(0), edge_index.size(1), weight.size(2)
 
     x = x.unsqueeze(-1) if x.dim() == 1 else x
     pseudo = pseudo.unsqueeze(-1) if pseudo.dim() == 1 else pseudo
 
-    # Weight gathered features based on B-spline bases and trainable weights.
+    # Convolve over each node.
     output = spline_weighting(x[edge_index[1]], pseudo, weight, kernel_size,
                               is_open_spline, degree)
 
@@ -57,4 +30,20 @@ def basic_spline_conv(x, edge_index, pseudo, weight, kernel_size,
     zero = x.new(n, m_out) if torch.is_tensor(x) else Var(x.data.new(n, m_out))
     output = zero.fill_(0).scatter_add_(0, row, output)
 
+    # Compute degree.
+    degree = x.new() if torch.is_tensor(x) else x.data.new()
+    degree = node_degree(edge_index, n, out=degree)
+
+    # Normalize output by node degree.
+    degree = degree.unsqueeze(-1).clamp_(min=1)
+    output /= degree if torch.is_tensor(x) else Var(degree)
+
+    # Weight root node separately (if wished).
+    if root_weight is not None:
+        output += torch.mm(x, root_weight)
+
+    # Add bias (if wished).
+    if bias is not None:
+        output += bias
+
     return output