[GitHub] [incubator-mxnet] drivanov commented on a change in pull request #16446: Aggregated zero grad

[GitBox <gi...@apache.org>]

drivanov commented on a change in pull request #16446: Aggregated zero grad
URL: https://github.com/apache/incubator-mxnet/pull/16446#discussion_r334556951
 
 

 ##########
 File path: src/operator/contrib/reset_arrays-inl.h
 ##########
 @@ -0,0 +1,99 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied.  See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+/*!
+ *  Copyright (c) 2019 by Contributors
+ * \file reset_arrays-inl.h
+ * \brief setting all array element values to zeros
+ * \author Moises Hernandez-Fernandez, Andrei Ivanov
+ */
+
+#ifndef MXNET_OPERATOR_CONTRIB_RESET_ARRAYS_INL_H_
+#define MXNET_OPERATOR_CONTRIB_RESET_ARRAYS_INL_H_
+
+#include <mxnet/operator_util.h>
+#include <vector>
+#include <limits>
+#include <algorithm>
+#include "../elemwise_op_common.h"
+#include "../mshadow_op.h"
+#include "../mxnet_op.h"
+#include "../tensor/init_op.h"
+
+namespace mxnet {
+namespace op {
+
+struct ResetArraysParam : public dmlc::Parameter<ResetArraysParam> {
+  int num_arrays;
+  DMLC_DECLARE_PARAMETER(ResetArraysParam) {
+    DMLC_DECLARE_FIELD(num_arrays)
+    .describe("number of input arrays.");
+  }
+};
+inline bool ResetArraysShape(const NodeAttrs& attrs,
+                            std::vector<mxnet::TShape>* in_shape,
+                            std::vector<mxnet::TShape>* out_shape) {
+  const auto& p = dmlc::get<ResetArraysParam>(attrs.parsed);
+  CHECK_EQ(in_shape->size(), p.num_arrays);
+  for (auto s : *in_shape) {
+    if (s.ndim() == 0)
+      return false;
+  }
+  return true;
+}
+
+inline bool ResetArraysType(const NodeAttrs& attrs,
+                           std::vector<int>* in_type,
+                           std::vector<int>* out_type) {
+  const auto& param_ = dmlc::get<ResetArraysParam>(attrs.parsed);
+  CHECK_EQ(in_type->size(), param_.num_arrays);
+  const int dtype = (*in_type)[0];
+  CHECK_NE(dtype, -1) << "First input must have specified type";
+  for (size_t i = 0; i < in_type->size(); ++i) {
+    const auto currType = (*in_type)[i];
+    if (currType == -1)
+      (*in_type)[i] = dtype;
 
 Review comment:
   I completely agree with you on both issues. But, it seems that Moises (who was the author of this code) was trying to follow common pattern which is used in 25 other MxNet operators (to find all of them, I searched for"First input must have specified type"). Do you think that for this operator we should break we should break this pattern? 

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services