[bugfix] fix the bug of DotGatConv when compute the edge_softmax (#3272)

Co-authored-by: Zihao Ye <expye@outlook.com>

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

@@ -208,7 +208,7 @@ class DotGatConv(nn.Module):
         graph.apply_edges(fn.u_dot_v('ft', 'ft', 'a'))
 
         # Step 2. edge softmax to compute attention scores
-        graph.edata['sa'] = edge_softmax(graph, graph.edata['a'])/(self._out_feats**0.5)
+        graph.edata['sa'] = edge_softmax(graph, graph.edata['a'] / self._out_feats**0.5)
 
         # Step 3. Broadcast softmax value to each edge, and aggregate dst node
         graph.update_all(fn.u_mul_e('ft', 'sa', 'attn'), fn.sum('attn', 'agg_u'))