From 4d649893ecf95042c18c898264d4d4a200a8a7f1 Mon Sep 17 00:00:00 2001
From: ziyoujiyi <73728031+ziyoujiyi@users.noreply.github.com>
Date: Fri, 17 Jun 2022 14:48:57 +0800
Subject: [PATCH] bug fix (#43526)

* back fl

* delete ssl cert

* .

* make warning

* .

* unittest paral degree

* solve unittest

* heter & multi cloud commm ready

* .

* .

* fl-ps v1.0

* .

* support N + N mode

* .

* .

* .

* .

* delete print

* .

* .

* .

* .

* fix bug

* .

* .
---
 paddle/fluid/framework/heter_pipeline_trainer.cc |  2 +-
 .../paddle/distributed/passes/ps_trainer_pass.py |  4 ++--
 python/paddle/distributed/ps/the_one_ps.py       | 16 +++-------------
 3 files changed, 6 insertions(+), 16 deletions(-)
 mode change 100644 => 100755 paddle/fluid/framework/heter_pipeline_trainer.cc

diff --git a/paddle/fluid/framework/heter_pipeline_trainer.cc b/paddle/fluid/framework/heter_pipeline_trainer.cc
old mode 100644
new mode 100755
index dc99885811..98860cfbb0
--- a/paddle/fluid/framework/heter_pipeline_trainer.cc
+++ b/paddle/fluid/framework/heter_pipeline_trainer.cc
@@ -333,5 +333,5 @@ Scope* HeterPipelineTrainer::GetWorkerScope(int thread_id) {
 }
 
 }  // end namespace framework
-}  // namespace paddle
+}  // end namespace paddle
 #endif
diff --git a/python/paddle/distributed/passes/ps_trainer_pass.py b/python/paddle/distributed/passes/ps_trainer_pass.py
index 4a015fea30..80012e7428 100755
--- a/python/paddle/distributed/passes/ps_trainer_pass.py
+++ b/python/paddle/distributed/passes/ps_trainer_pass.py
@@ -434,8 +434,8 @@ class DistributedOpsPass(PassBase):
             if op.type in SPARSE_OP_TYPE_DICT.keys() \
                     and op.attr('remote_prefetch') is True:
                 param_name = op.input(SPARSE_OP_TYPE_DICT[op.type])[0]
-                if attrs['is_heter_ps_mode']:
-                    # trick for matchnet, need to modify
+                if attrs['is_heter_ps_mode'] and not attrs['is_fl_ps_mode']:
+                    # TODO: trick for matchnet, need to modify for heter_ps
                     param_name += op.input("Ids")[0][0]
                 ops = pull_sparse_ops.get(param_name, [])
                 ops.append(op)
diff --git a/python/paddle/distributed/ps/the_one_ps.py b/python/paddle/distributed/ps/the_one_ps.py
index 0836e91c30..a199901011 100755
--- a/python/paddle/distributed/ps/the_one_ps.py
+++ b/python/paddle/distributed/ps/the_one_ps.py
@@ -1015,14 +1015,8 @@ class TheOnePSRuntime(RuntimeBase):
 
         is_test = bool(int(os.getenv("TEST_MODE", "0")))
 
-        # for GEO
-        if self.role_maker._is_first_worker() and self.is_heter_ps_mode:
-            # for ps-heter mode load all parameters on first_worker
-            init_params = get_the_one_recv_context(self.context,
-                                                   split_dense_table=True,
-                                                   use_origin_program=True)
-        else:
-            init_params = dense_map
+        # for GEO & heter_ps
+        init_params = dense_map
 
         # if not is_test:
         #     self._communicator.init_params(init_params)
@@ -1053,11 +1047,7 @@ class TheOnePSRuntime(RuntimeBase):
             fleet.util.barrier()  # 保证 0 号 worker 参数 push_dense_param over
 
         if not self.context['use_ps_gpu']:
-            if self.is_heter_ps_mode == True and not self.role_maker._is_first_worker(
-            ):
-                self._communicator.pull_dense(init_params)
-            else:
-                self._pull_all_dense(scopes, send_ctx, dense_map)
+            self._pull_all_dense(scopes, send_ctx, dense_map)
         fleet.util.barrier()
 
         if self.context[
-- 
GitLab