fix black

gallery/how_to/tune_with_autoscheduler/tune_sparse_x86.py

@@ -239,6 +239,7 @@ def apply_func(search_policy, state, stage_id):
 # After some measurement trials, we can load the best schedule from the log
 # file and apply it.
 
+
 def tune_and_evaluate(tune_option, search_policy):
     # Run auto-tuning (search)
     task.tune(tune_option, search_policy)
@@ -275,11 +276,14 @@ def tune_and_evaluate(tune_option, search_policy):
     print(
         "Execution time of this operator: %.3f ms"
         % (
-            np.median(evaluator(X_tvm, W_data_tvm, W_indices_tvm, W_indptr_tvm, B_tvm, Y_tvm).results)
+            np.median(
+                evaluator(X_tvm, W_data_tvm, W_indices_tvm, W_indptr_tvm, B_tvm, Y_tvm).results
+            )
             * 1000
         )
     )
 
+
 # Notice: We do not run the tuning in our webpage server since it takes too long.
 # Uncomment the following line to run it by yourself.
 # tune_and_evaluate(tune_option, search_policy)

gallery/how_to/work_with_relay/build_gcn.py

@@ -115,7 +115,7 @@ def evaluate(g, logits):
 dgl_g = dataset[0]
 num_layers = 1
 num_hidden = 16
-features = dgl_g.ndata['feat']
+features = dgl_g.ndata["feat"]
 infeat_dim = features.shape[1]
 num_classes = dataset.num_classes
 
@@ -232,9 +232,7 @@ def GraphConv(layer_name, input_dim, output_dim, adj, input, norm=None, bias=Tru
 
 def prepare_params(g):
     params = {}
-    params["infeats"] = g.ndata['feat'].numpy().astype(
-        "float32"
-    )  # Only support float32 as feature for now
+    params["infeats"] = g.ndata["feat"].numpy().astype("float32")
 
     # Generate adjacency matrix
     nx_graph = dgl.to_networkx(g)