fix conflict of inferne partial feed with gpu parallel ssa graph executor, test=develop (#23400)

paddle/fluid/framework/details/multi_devices_helper.cc

@@ -105,38 +105,36 @@ static size_t GetUniqueDeviceIdOfOp(const details::OpHandleBase &op) {
   return dev_idx;
 }
 
-/**
- * This function tries to separate the original graph into multiple graphs, in
- * which each graph would only run on single device. This is usually used to
- * separate a data-parallel inference graph to multiple graphs on each device.
- *
- * The graph can be separated into multiple single device graphs if and only if:
- *
- *  - the graph does not contain any ops related to multi-devices communication,
- *    such as allreduce, send, recv, sync_batch_norm, etc.
- *
- *  - ops on different devices do not depend on each other. That is to say, the
- *    graph has several disconnected sub-graphs.
- */
-std::vector<std::unique_ptr<ir::Graph>> TrySeparateToMultipleSingleDeviceGraphs(
-    ir::Graph *graph) {
+static bool IsDataParallelInferenceGraphImpl(
+    const ir::Graph &graph,
+    std::unordered_map<details::OpHandleBase *, size_t> *p_op_to_dev_idx,
+    size_t *p_place_num) {
+  auto &place_num = *p_place_num;
+  auto &op_to_dev_idx = *p_op_to_dev_idx;
+
+  auto clear_result = [&] {
+    place_num = 0;
+    op_to_dev_idx.clear();
+    return false;
+  };
+
+  clear_result();
+
   // If sub-block contains multi-devices ops, we cannot separate
-  if (ContainMultiDeviceOp(graph->OriginProgram(), 1)) {
-    return {};
+  if (ContainMultiDeviceOp(graph.OriginProgram(), 1)) {
+    return clear_result();
   }
 
-  size_t place_num = 0;
-  auto op_handles = ir::FilterByNodeWrapper<OpHandleBase>(*graph);
+  auto op_handles = ir::FilterByNodeWrapper<OpHandleBase>(graph);
   if (op_handles.empty()) {
-    return {};
+    return clear_result();
   }
 
-  std::unordered_map<details::OpHandleBase *, size_t> op_to_dev_idx;
   for (auto &op : op_handles) {
     auto dev_idx = GetUniqueDeviceIdOfOp(*op);
     if (dev_idx == kUndefinedDevIdx) {
       VLOG(10) << "Op " << op->Name() << " is not determined";
-      return {};
+      return clear_result();
     }
     place_num = std::max(place_num, dev_idx + 1);
     op_to_dev_idx[op] = dev_idx;
@@ -148,7 +146,7 @@ std::vector<std::unique_ptr<ir::Graph>> TrySeparateToMultipleSingleDeviceGraphs(
       if (in_var->GeneratedOp()) {
         auto iter = op_to_dev_idx.find(in_var->GeneratedOp());
         if (iter == op_to_dev_idx.end() || iter->second != dev_idx) {
-          return {};
+          return clear_result();
         }
       }
     }
@@ -157,7 +155,7 @@ std::vector<std::unique_ptr<ir::Graph>> TrySeparateToMultipleSingleDeviceGraphs(
       for (auto &pending_op : out_var->PendingOps()) {
         auto iter = op_to_dev_idx.find(pending_op);
         if (iter == op_to_dev_idx.end() || iter->second != dev_idx) {
-          return {};
+          return clear_result();
         }
       }
     }
@@ -171,6 +169,36 @@ std::vector<std::unique_ptr<ir::Graph>> TrySeparateToMultipleSingleDeviceGraphs(
           "issue at https://github.com/PaddlePaddle/Paddle/issues/new. And "
           "we will resolve it with high priority."));
 
+  return true;
+}
+
+bool IsDataParallelInferenceGraph(const ir::Graph &graph) {
+  size_t place_num;
+  std::unordered_map<details::OpHandleBase *, size_t> op_to_dev_idx;
+  return IsDataParallelInferenceGraphImpl(graph, &op_to_dev_idx, &place_num);
+}
+
+/**
+ * This function tries to separate the original graph into multiple graphs, in
+ * which each graph would only run on single device. This is usually used to
+ * separate a data-parallel inference graph to multiple graphs on each device.
+ *
+ * The graph can be separated into multiple single device graphs if and only if:
+ *
+ *  - the graph does not contain any ops related to multi-devices communication,
+ *    such as allreduce, send, recv, sync_batch_norm, etc.
+ *
+ *  - ops on different devices do not depend on each other. That is to say, the
+ *    graph has several disconnected sub-graphs.
+ */
+std::vector<std::unique_ptr<ir::Graph>> TrySeparateToMultipleSingleDeviceGraphs(
+    ir::Graph *graph) {
+  size_t place_num;
+  std::unordered_map<details::OpHandleBase *, size_t> op_to_dev_idx;
+  if (!IsDataParallelInferenceGraphImpl(*graph, &op_to_dev_idx, &place_num)) {
+    return {};
+  }
+
   if (place_num == 1) {
     return {};
   }
@@ -182,8 +210,10 @@ std::vector<std::unique_ptr<ir::Graph>> TrySeparateToMultipleSingleDeviceGraphs(
     g->Set(kGraphDepVars, new GraphDepVars());
   }
 
-  for (auto &op : op_handles) {
-    auto dev_idx = op_to_dev_idx.at(op);
+  for (auto &pair : op_to_dev_idx) {
+    auto *op = pair.first;
+    auto dev_idx = pair.second;
+
     auto *ret_graph = graphs[dev_idx].get();
     auto &ret_vars = ret_graph->Get<GraphVars>(kGraphVars)[0];
     auto &ret_dummy_vars = ret_graph->Get<GraphDepVars>(kGraphDepVars);

paddle/fluid/framework/details/multi_devices_helper.h

@@ -101,6 +101,8 @@ inline std::vector<std::string> GetOpRoleVarsOrEmpty(const OpDesc &op) {
   return boost::get<std::vector<std::string>>(iter->second);
 }
 
+bool IsDataParallelInferenceGraph(const ir::Graph &graph);
+
 std::vector<std::unique_ptr<ir::Graph>> TrySeparateToMultipleSingleDeviceGraphs(
     ir::Graph *graph);
 

paddle/fluid/framework/parallel_executor.cc

@@ -647,11 +647,22 @@ ParallelExecutor::ParallelExecutor(const std::vector<platform::Place> &places,
 #ifdef PADDLE_WITH_CUDA
     // TODO(Yancey1989): Remove passing in the main_program when
     // allreduce_seq_pass doesn't need it as the attr.
+    bool is_inference = details::IsDataParallelInferenceGraph(*graph);
+    bool has_drop_last_read_op = details::HasDropLastReadOp(*graph);
+
     auto *pg_exe = new details::ParallelSSAGraphExecutor(
         exec_strategy, member_->local_scopes_, member_->local_exec_scopes_,
         member_->places_, graph);
     final_graphs = pg_exe->Graphs();
     member_->executor_.reset(pg_exe);
+
+    if (is_inference && member_->places_.size() > 1) {
+      member_->inference_executor_ = pg_exe;
+      if (!has_drop_last_read_op) {
+        VLOG(5) << "Enable partial feed support in inference phase";
+        pg_exe->EnablePartialFeedSupport();
+      }
+    }
 #else
     PADDLE_THROW(
         "Paddle should be compiled with CUDA for ParallelGraph Execution.");

python/paddle/fluid/tests/unittests/CMakeLists.txt

@@ -357,6 +357,7 @@ set_tests_properties(test_parallel_executor_test_while_train test_parallel_execu
         test_optimizer_in_control_flow test_dataloader_keep_order
         test_dataloader_unkeep_order
         test_parallel_executor_inference_feed_partial_data
+        test_parallel_ssa_graph_inference_feed_partial_data
         test_fetch_unmerged
         test_buffer_shared_memory_reuse_pass PROPERTIES LABELS "RUN_TYPE=DIST")
 

python/paddle/fluid/tests/unittests/test_parallel_ssa_graph_inference_feed_partial_data.py

@@ -0,0 +1,23 @@
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed 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.
+
+import paddle.fluid as fluid
+import unittest
+
+fluid.core.globals()['FLAGS_enable_parallel_graph'] = 1
+
+from test_parallel_executor_inference_feed_partial_data import *
+
+if __name__ == '__main__':
+    unittest.main()