[incubator-mxnet] 01/03: no bulk exec for aggregate

[cj...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

cjolivier01 pushed a commit to branch cython
in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git

commit 306626c3fd6e9124f436ae85fc1fc192912302ad
Author: Olivier <co...@amazon.com>
AuthorDate: Mon Mar 12 13:41:47 2018 -0700

    no bulk exec for aggregate
---
 src/executor/graph_executor.cc | 13 +++++++++----
 src/imperative/imperative.cc   |  7 ++++++-
 2 files changed, 15 insertions(+), 5 deletions(-)

diff --git a/src/executor/graph_executor.cc b/src/executor/graph_executor.cc
index 7d31a31..8c0ac52 100644
--- a/src/executor/graph_executor.cc
+++ b/src/executor/graph_executor.cc
@@ -1346,11 +1346,12 @@ void GraphExecutor::InitOpSegs() {
   if (monitor_callback_) return;
 
   // Generate segments based on the graph structure
-  bool prefer_bulk_exec_inference = dmlc::GetEnv("MXNET_EXEC_BULK_EXEC_INFERENCE", true);
+  const bool prefer_bulk_exec_inference = dmlc::GetEnv("MXNET_EXEC_BULK_EXEC_INFERENCE", true);
+
   // Whether to perform bulk exec for training
-  bool prefer_bulk_exec = dmlc::GetEnv("MXNET_EXEC_BULK_EXEC_TRAIN", 1);
+  const bool prefer_bulk_exec = dmlc::GetEnv("MXNET_EXEC_BULK_EXEC_TRAIN", 1) != 0;
 
-  bool is_training = num_forward_nodes_ != total_num_nodes;
+  const bool is_training = num_forward_nodes_ != total_num_nodes;
 
   if (prefer_bulk_exec  && is_training) {
     this->BulkTrainingOpSegs(total_num_nodes);
@@ -1365,7 +1366,11 @@ void GraphExecutor::InitOpSegs() {
 
 void GraphExecutor::BulkTrainingOpSegs(size_t total_num_nodes) {
   // The maximum number of node in a segment executed in bulk
-  size_t num_nodes_threshold = dmlc::GetEnv("MXNET_EXEC_BULK_EXEC_MAX_NODE_TRAIN", 15);
+
+  const size_t num_nodes_threshold = dmlc::GetEnv("MXNET_EXEC_BULK_EXEC_MAX_NODE_TRAIN", 15U);
+
+//  const size_t num_nodes_threshold = profiler::Profiler::Get()->AggregateEnabled() ? 1 :
+//                                     dmlc::GetEnv("MXNET_EXEC_BULK_EXEC_MAX_NODE_TRAIN", 15U);
 
   // create forward segments for training
   size_t topo_start = 0;
diff --git a/src/imperative/imperative.cc b/src/imperative/imperative.cc
index fbbaf82..a915255 100644
--- a/src/imperative/imperative.cc
+++ b/src/imperative/imperative.cc
@@ -18,6 +18,7 @@
  */
 #include <unordered_set>
 #include <iostream>
+#include "../profiler/profiler.h"
 #include "./imperative_utils.h"
 
 namespace mxnet {
@@ -579,7 +580,11 @@ std::vector<NDArray*> Imperative::Backward(
 
   bool prev_recording = set_is_recording(create_graph);
   bool prev_training = set_is_training(is_train);
-  int prev_bulk_size = Engine::Get()->set_bulk_size(backward_bulk_size_);
+
+//  const int prev_bulk_size = Engine::Get()->set_bulk_size(
+//    profiler::Profiler::Get()->AggregateEnabled() ? 1 : backward_bulk_size_);
+
+  const int prev_bulk_size = Engine::Get()->set_bulk_size(backward_bulk_size_);
 
   RunGraph(retain_graph, idx, arrays, num_forward_nodes, idx.num_nodes(),
            std::move(array_reqs), std::move(ref_count), &states, dispatch_modes);

-- 
To stop receiving notification emails like this one, please contact
cjolivier01@apache.org.