From 542b52fac32b2e8b494be8145c686ba202ffa729 Mon Sep 17 00:00:00 2001 From: Qiao Longfei Date: Mon, 25 Mar 2019 10:03:57 +0800 Subject: [PATCH] fix trainer_id --- paddle/fluid/operators/distributed/rpc_client.h | 1 + python/paddle/fluid/transpiler/distribute_transpiler.py | 2 ++ 2 files changed, 3 insertions(+) diff --git a/paddle/fluid/operators/distributed/rpc_client.h b/paddle/fluid/operators/distributed/rpc_client.h index f893510ba..86d483cdd 100644 --- a/paddle/fluid/operators/distributed/rpc_client.h +++ b/paddle/fluid/operators/distributed/rpc_client.h @@ -97,6 +97,7 @@ class RPCClient { // Init is called by GetInstance. template static void Init(int trainer_id) { + VLOG(0) << "init rpc client with trainer_id " << trainer_id; trainer_id_ = trainer_id; if (rpc_client_.get() == nullptr) { rpc_client_.reset(new T()); diff --git a/python/paddle/fluid/transpiler/distribute_transpiler.py b/python/paddle/fluid/transpiler/distribute_transpiler.py index 0c3290cdc..643e98613 100644 --- a/python/paddle/fluid/transpiler/distribute_transpiler.py +++ b/python/paddle/fluid/transpiler/distribute_transpiler.py @@ -658,6 +658,7 @@ class DistributeTranspiler(object): outputs={"Out": splited_var}, attrs={ "epmap": eps, + "trainer_id": self.trainer_id, RPC_OP_ROLE_ATTR_NAME: RPC_OP_ROLE_ATTR_VALUE }) @@ -669,6 +670,7 @@ class DistributeTranspiler(object): outputs={"Out": fetch_barrier_out}, attrs={ "endpoints": self.pserver_endpoints, + "trainer_id": self.trainer_id, RPC_OP_ROLE_ATTR_NAME: RPC_OP_ROLE_ATTR_VALUE }) -- GitLab