make shape inference of BatchNorm layout neutral (dmlc#301)

* make shape inference of BatchNorm layout neutral

* refactor to use the axis variable to do BatchNorm shape inference

* refactor to use the axis variable to do BatchNorm shape inference

* add unittest to the axis param for batch norm shape inference

src/top/nn/nn.cc

@@ -117,12 +117,15 @@ DMLC_REGISTER_PARAMETER(BatchNormParam);
 inline bool BatchNormInferShape(const nnvm::NodeAttrs& attrs,
                                 std::vector<TShape>* in_shape,
                                 std::vector<TShape>* out_shape) {
+  const BatchNormParam& param = nnvm::get<BatchNormParam>(attrs.parsed);
   CHECK_EQ(in_shape->size(), 5U)
       << "Input:[data, gamma, beta, moving_mean, moving_var]";
   CHECK_EQ(out_shape->size(), 3U);
   const TShape &dshape = in_shape->at(0);
   if (dshape.ndim() == 0) return false;
-  TShape bshape({dshape[1]});
+  CHECK((size_t)param.axis < dshape.Size());
+
+  TShape bshape({dshape[param.axis]});
   NNVM_ASSIGN_INPUT_SHAPE(attrs, *in_shape, 1, bshape);
   NNVM_ASSIGN_INPUT_SHAPE(attrs, *in_shape, 2, bshape);
   NNVM_ASSIGN_INPUT_SHAPE(attrs, *in_shape, 3, bshape);

tests/python/unittest/test_infer_shape.py

@@ -62,6 +62,22 @@ def test_batchnorm():
     sdict = infer_shape(y)
     assert(sdict["bn_gamma"][0] == [20])
 
+    x = sym.Variable("x", shape=(10, 20, 30, 40))
+    y = sym.batch_norm(data=x, axis=0, epsilon=2e-5, name='bn')
+    sdict = infer_shape(y)
+    assert(sdict['bn_moving_var'][0] == [10])
+
+    y = sym.batch_norm(data=x, axis=1, epsilon=2e-5, name='bn')
+    sdict = infer_shape(y)
+    assert(sdict['bn_gamma'][0] == [20])
+
+    y = sym.batch_norm(data=x, axis=2, epsilon=2e-5, name='bn')
+    sdict = infer_shape(y)
+    assert(sdict['bn_beta'][0] == [30])
+
+    y = sym.batch_norm(data=x, axis=3, epsilon=2e-5, name='bn')
+    sdict = infer_shape(y)
+    assert(sdict['bn_moving_mean'][0] == [40])
 
 def test_flatten():
     x = sym.Variable("x", shape=(10, 20, 10))