[WIP] Misc Fix (#2359)

* Update

* Update

* Update

docs/source/api/python/dgl.function.rst

@@ -112,8 +112,6 @@ Here is a cheatsheet of all the DGL built-in functions.
 |                         +-----------------------------------------------------------------+-----------------------+
 |                         | ``sum``                                                         |                       |
 |                         +-----------------------------------------------------------------+-----------------------+
-|                         | ``prod``                                                        |                       |
-|                         +-----------------------------------------------------------------+-----------------------+
 |                         | ``mean``                                                        |                       |
 +-------------------------+-----------------------------------------------------------------+-----------------------+
 
@@ -168,5 +166,4 @@ Reduce functions
     sum
     max
     min
-    prod
     mean

docs/source/guide/message-api.rst

@@ -58,6 +58,8 @@ to the Reduce UDF that sums up the message ``m``:
     def reduce_func(nodes):
          return {'h': torch.sum(nodes.mailbox['m'], dim=1)}
 
+For advanced usage of UDF, see :ref:`apiudf`.
+
 It is also possible to invoke only edge-wise computation by :meth:`~dgl.DGLGraph.apply_edges`
 without invoking message passing. :meth:`~dgl.DGLGraph.apply_edges` takes a message function
 for parameter and by default updates the features of all edges. For example:

docs/source/guide_cn/message-api.rst

@@ -43,6 +43,8 @@ DGL支持内置的聚合函数 ``sum``、 ``max``、 ``min`` 和 ``mean`` 操作
     def reduce_func(nodes):
          return {'h': torch.sum(nodes.mailbox['m'], dim=1)}
 
+关于用户定义函数的进阶用法，参见 :ref:`apiudf`。
+
 在DGL中，也可以在不涉及消息传递的情况下，通过 :meth:`~dgl.DGLGraph.apply_edges` 单独调用逐边计算。
 :meth:`~dgl.DGLGraph.apply_edges` 的参数是一个消息函数。并且在默认情况下，这个接口将更新所有的边。例如：
 

python/dgl/nn/pytorch/conv/nnconv.py

@@ -151,8 +151,9 @@ class NNConv(nn.Module):
             is the number of nodes of the graph and :math:`D_{in}` is the
             input feature size.
         efeat : torch.Tensor
-            The edge feature of shape :math:`(N, *)`, should fit the input
-            shape requirement of ``edge_func``.
+            The edge feature of shape :math:`(E, *)`, which should fit the input
+            shape requirement of ``edge_func``. :math:`E` is the number of edges
+            of the graph.
 
         Returns
         -------