add scheduler and update demos

docs/requirements.txt

 sphinx==2.1.0
 mistune
 sphinx_rtd_theme
-paddlepaddle
+paddlepaddle>=1.6

paddle_fl/core/__init__.py

@@ -17,3 +17,6 @@ from .master.fl_job import FLRunTimeJob
 from .master.job_generator import JobGenerator
 from .strategy.fl_strategy_base import DPSGDStrategy
 from .strategy.fl_strategy_base import FedAvgStrategy
+from .scheduler.agent_master import FLServerAgent
+from .scheduler.agent_master import FLWorkerAgent
+from .scheduler.agent_master import FLScheduler

paddle_fl/core/master/fl_job.py

@@ -176,6 +176,7 @@ class FLRunTimeJob(FLJobBase):
         self._server_main_program = None
         self._feed_names = None
         self._target_names = None
+        self._scheduler_ep = None	
 
     def _load_strategy(self, input_file):
         import pickle

paddle_fl/core/scheduler/__init__.py

+#   Copyright (c) 2019 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.

paddle_fl/core/scheduler/agent_master.py

+import zmq
+import time
+import random
+
+def recv_and_parse_kv(socket):
+    message = socket.recv()
+    group = message.split("\t")
+    if group[0] == "alive":
+        return group[0], "0"
+    else:
+        return group[0], group[1]
+
+WORKER_EP = "WORKER_EP"
+SERVER_EP = "SERVER_EP"
+
+class FLServerAgent(object):
+    def __init__(self, scheduler_ep, current_ep):
+        self.scheduler_ep = scheduler_ep
+        self.context = zmq.Context()
+        self.socket = self.context.socket(zmq.REQ)
+        self.socket.connect("tcp://127.0.0.1:9091")
+        self.current_ep = current_ep
+
+    def connect_scheduler(self):
+        self.socket.send("SERVER_EP\t{}".format(self.current_ep))
+        self.socket.recv()
+
+
+class FLWorkerAgent(object):
+    def __init__(self, scheduler_ep, current_ep):
+        self.scheduler_ep = scheduler_ep
+        self.context = zmq.Context()
+        self.socket = self.context.socket(zmq.REQ)
+        self.socket.connect("tcp://127.0.0.1:9091")
+        self.current_ep = current_ep
+
+    def connect_scheduler(self):
+        self.socket.send("WORKER_EP\t{}".format(self.current_ep))
+        self.socket.recv()
+
+    def finish_training(self):
+        self.socket.send("FINISH\t{}".format(self.current_ep))
+        key, value = recv_and_parse_kv(self.socket)
+        if key == "WAIT":
+            time.sleep(3)
+	    return True
+	return False
+
+    def can_join_training(self):
+        self.socket.send("JOIN\t{}".format(self.current_ep))
+        key, value = recv_and_parse_kv(self.socket)
+
+        if key == "ACCEPT":
+            return True
+        elif key == "REJECT":
+            return False
+        return False
+
+
+
+class FLScheduler(object):
+    def __init__(self, worker_num, server_num, port=9091):
+        self.context = zmq.Context()
+        self.socket = self.context.socket(zmq.REP)
+        self.socket.bind("tcp://*:{}".format(port))
+        self.worker_num = worker_num
+        self.server_num = server_num
+        self.sample_worker_num = 0
+        self.fl_workers = []
+        self.fl_servers = []
+
+    def set_sample_worker_num(self, sample_worker_num=0):
+        if sample_worker_num == 0:
+            self.sample_worker_num = int(self.worker_num * 0.1)
+        else:
+            self.sample_worker_num = sample_worker_num
+
+    def init_env(self):
+        ready = False
+        while not ready:
+            key, value = recv_and_parse_kv(self.socket)
+            if key == WORKER_EP:
+                self.fl_workers.append(value)
+                self.socket.send("INIT\t{}".format(value))
+            elif key == SERVER_EP:
+                self.fl_servers.append(value)
+                self.socket.send("INIT\t{}".format(value))
+            else:
+                time.sleep(3)
+                self.socket.send("REJECT\t0")
+            if len(self.fl_workers) == self.worker_num and \
+               len(self.fl_servers) == self.server_num:
+                ready = True
+
+    def start_fl_training(self):
+        # loop until training is done
+	loop = 0
+        while True:
+	    if loop <= 1:
+		print(loop)
+            random.shuffle(self.fl_workers)
+            worker_dict = {}
+            for worker in self.fl_workers[:self.sample_worker_num]:
+                worker_dict[worker] = 0
+
+            ready_workers = []
+            all_ready_to_train = False
+            while not all_ready_to_train:
+                key, value = recv_and_parse_kv(self.socket)
+                if key == "JOIN":
+                    if value in worker_dict:
+                        if worker_dict[value] == 0:
+                            ready_workers.append(value)
+                            worker_dict[value] = 1
+                            self.socket.send("ACCEPT\t0")
+                            continue
+                    else:
+			if value not in ready_workers:
+                    		ready_workers.append(value)
+                self.socket.send("REJECT\t0")
+                if len(ready_workers) == len(self.fl_workers):
+                    all_ready_to_train = True
+
+            all_finish_training = False
+            finish_training_dict = {}
+            while not all_finish_training:
+                key, value = recv_and_parse_kv(self.socket)
+		if key == "FINISH":
+                    finish_training_dict[value] = 1
+                    self.socket.send("WAIT\t0")
+                else:
+                    self.socket.send("REJECT\t0")
+                if len(finish_training_dict) == len(worker_dict):
+                    all_finish_training = True
+            time.sleep(5)
+	    loop += 1

paddle_fl/core/server/fl_server.py

@@ -12,12 +12,15 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import paddle.fluid as fluid
+from paddle_fl.core.scheduler.agent_master import FLServerAgent
 
 class FLServer(object):
 
     def __init__(self):
         self._startup_program = None
         self._main_program = None
+	self._scheduler_ep = None
+	self._current_ep = None
 
     def set_server_job(self, job):
         # need to parse startup and main program in job
@@ -25,9 +28,12 @@ class FLServer(object):
         # need to parse master endpoint
         self._startup_program = job._server_startup_program
         self._main_program = job._server_main_program
+	self._scheduler_ep = job._scheduler_ep
+	self._current_ep = None
 
     def start(self):
+	self.agent = FLServerAgent(self._scheduler_ep, self._current_ep)
+	self.agent.connect_scheduler()
         exe = fluid.Executor(fluid.CPUPlace())
         exe.run(self._startup_program)
         exe.run(self._main_program)
-

paddle_fl/core/strategy/__init__.py

+#   Copyright (c) 2019 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.

paddle_fl/core/trainer/fl_trainer.py

@@ -13,6 +13,7 @@
 # limitations under the License.
 import paddle.fluid as fluid
 import logging
+from paddle_fl.core.scheduler.agent_master import FLWorkerAgent
 
 class FLTrainerFactory(object):
     def __init__(self):
@@ -44,17 +45,24 @@ class FLTrainer(object):
         self._step = job._strategy._inner_step
         self._feed_names = job._feed_names
         self._target_names = job._target_names
+        self._scheduler_ep = job._scheduler_ep
+	self._current_ep = None
+	self.cur_step = 0
 
     def start(self):
+        #current_ep = "to be added"
+        self.agent = FLWorkerAgent(self._scheduler_ep, self._current_ep)
+        self.agent.connect_scheduler()
         self.exe = fluid.Executor(fluid.CPUPlace())
         self.exe.run(self._startup_program)
 
     def run(self, feed, fetch):
         self._logger.debug("begin to run")
         self.exe.run(self._main_program,
-                     feed=feed,
-                     fetch_list=fetch)
+                      feed=feed,
+                      fetch_list=fetch)
         self._logger.debug("end to run current batch")
+	self.cur_step += 1
 
     def save_inference_program(self, output_folder):
         target_vars = []
@@ -73,7 +81,15 @@ class FLTrainer(object):
         # ask for termination with master endpoint
         # currently not open sourced, will release the code later
         # TODO(guru4elephant): add connection with master
-        return False
+	if self.cur_step != 0:
+		while not self.agent.finish_training():
+			print('wait others finish')
+			continue
+        while not self.agent.can_join_training():
+		print("wait permit")
+		continue    
+	print("ready to train")
+	return False
 
 class FedAvgTrainer(FLTrainer):
     def __init__(self):
@@ -81,9 +97,11 @@ class FedAvgTrainer(FLTrainer):
         pass
 
     def start(self):
+	#current_ep = "to be added"
+        self.agent = FLWorkerAgent(self._scheduler_ep, self._current_ep)
+	self.agent.connect_scheduler()
         self.exe = fluid.Executor(fluid.CPUPlace())
         self.exe.run(self._startup_program)
-        self.cur_step = 0
 
     def set_trainer_job(self, job):
         super(FedAvgTrainer, self).set_trainer_job(job)
@@ -108,7 +126,4 @@ class FedAvgTrainer(FLTrainer):
             self.exe.run(self._send_program)
         self.cur_step += 1
         return loss
-
-    def stop(self):
-        return False
         

paddle_fl/examples/ctr_demo/fl_master.py

@@ -38,7 +38,7 @@ job_generator.set_infer_feed_and_target_names(
     [x.name for x in inputs], [model.predict.name])
 
 build_strategy = FLStrategyFactory()
-build_strategy.fed_avg = True
+build_strategy.fed_avg = True 
 build_strategy.inner_step = 10
 strategy = build_strategy.create_fl_strategy()
 
@@ -47,5 +47,5 @@ strategy = build_strategy.create_fl_strategy()
 endpoints = ["127.0.0.1:8181"]
 output = "fl_job_config"
 job_generator.generate_fl_job(
-    strategy, server_endpoints=endpoints, worker_num=2, output=output)
+    strategy, server_endpoints=endpoints, worker_num=5, output=output)
 # fl_job_config will  be dispatched to workers

paddle_fl/examples/ctr_demo/fl_scheduler.py

+from paddle_fl.core.scheduler.agent_master import FLScheduler
+
+worker_num = 5
+server_num = 1
+scheduler = FLScheduler(worker_num,server_num)
+scheduler.set_sample_worker_num(5)
+scheduler.init_env()
+print("init env done.")
+scheduler.start_fl_training()

paddle_fl/examples/ctr_demo/fl_server.py

@@ -21,5 +21,8 @@ server_id = 0
 job_path = "fl_job_config"
 job = FLRunTimeJob()
 job.load_server_job(job_path, server_id)
+job._scheduler_ep = "127.0.0.1:9091"
 server.set_server_job(job)
+server._current_ep = "127.0.0.1:8181"
 server.start()
+print("connect")

paddle_fl/examples/ctr_demo/fl_trainer.py

@@ -3,6 +3,7 @@ from paddle_fl.core.master.fl_job import FLRunTimeJob
 import numpy as np
 import sys
 import logging
+import time
 logging.basicConfig(filename="test.log", filemode="w", format="%(asctime)s %(name)s:%(levelname)s:%(message)s", datefmt="%d-%M-%Y %H:%M:%S", level=logging.DEBUG)
 
 
@@ -18,15 +19,22 @@ trainer_id = int(sys.argv[1]) # trainer id for each guest
 job_path = "fl_job_config"
 job = FLRunTimeJob()
 job.load_trainer_job(job_path, trainer_id)
+job._scheduler_ep = "127.0.0.1:9091"
 trainer = FLTrainerFactory().create_fl_trainer(job)
+trainer._current_ep = "127.0.0.1:{}".format(9000+trainer_id)
 trainer.start()
 
+print(trainer._scheduler_ep, trainer._current_ep)
 output_folder = "fl_model"
 step_i = 0
 while not trainer.stop():
-    step_i += 1
     print("batch %d start train" % (step_i))
+    train_step = 0
     for data in reader():
-        trainer.run(feed=data, fetch=[])
+      	trainer.run(feed=data, fetch=[])
+	train_step += 1	
+	if train_step == trainer._step:
+		break
+    step_i += 1
     if step_i % 100 == 0:
         trainer.save_inference_program(output_folder)

paddle_fl/examples/ctr_demo/run.sh

@@ -2,8 +2,13 @@ unset http_proxy
 unset https_proxy
 python fl_master.py
 sleep 2
+python -u fl_scheduler.py > scheduler.log &
+sleep 5
 python -u fl_server.py >server0.log &
 sleep 2
-python -u fl_trainer.py 0 >trainer0.log &
-sleep 2
-python -u fl_trainer.py 1 >trainer1.log &
+for ((i=0;i<5;i++))
+do 
+    python -u fl_trainer.py $i >trainer$i.log &
+    sleep 2
+done
+	

paddle_fl/examples/dpsgd_demo/fl_master.py

@@ -49,5 +49,5 @@ strategy.sigma = CLIP * SIGMA
 endpoints = ["127.0.0.1:8181"]
 output = "fl_job_config"
 job_generator.generate_fl_job(
-    strategy, server_endpoints=endpoints, worker_num=2, output=output)
+    strategy, server_endpoints=endpoints, worker_num=4, output=output)
 # fl_job_config will  be dispatched to workers

paddle_fl/examples/dpsgd_demo/fl_scheduler.py

+from paddle_fl.core.scheduler.agent_master import FLScheduler
+
+worker_num = 4
+server_num = 1
+scheduler = FLScheduler(worker_num,server_num)
+scheduler.set_sample_worker_num(4)
+scheduler.init_env()
+print("init env done.")
+scheduler.start_fl_training()

paddle_fl/examples/dpsgd_demo/fl_server.py

@@ -21,5 +21,7 @@ server_id = 0
 job_path = "fl_job_config"
 job = FLRunTimeJob()
 job.load_server_job(job_path, server_id)
+job._scheduler_ep = "127.0.0.1:9091"
 server.set_server_job(job)
+server._current_ep = "127.0.0.1:8181"
 server.start()

paddle_fl/examples/dpsgd_demo/fl_trainer.py

@@ -13,7 +13,9 @@ trainer_id = int(sys.argv[1]) # trainer id for each guest
 job_path = "fl_job_config"
 job = FLRunTimeJob()
 job.load_trainer_job(job_path, trainer_id)
+job._scheduler_ep = "127.0.0.1:9091"
 trainer = FLTrainerFactory().create_fl_trainer(job)
+trainer._current_ep = "127.0.0.1:{}".format(9000+trainer_id)
 trainer.start()
 
 test_program = trainer._main_program.clone(for_test=True)

paddle_fl/examples/dpsgd_demo/run.sh

 python fl_master.py
 sleep 2
+python -u fl_scheduler.py >scheduler.log &
+sleep 2
 python -u fl_server.py >server0.log &
 sleep 2
 python -u fl_trainer.py 0 >trainer0.log &
 sleep 2
 python -u fl_trainer.py 1 >trainer1.log &
-
+sleep 2
+python -u fl_trainer.py 2 >trainer2.log &
+sleep 2
+python -u fl_trainer.py 3 >trainer3.log &

paddle_fl/examples/gru4rec_demo/fl_master.py

@@ -73,7 +73,7 @@ job_generator.set_infer_feed_and_target_names(
 
 build_strategy = FLStrategyFactory()
 build_strategy.fed_avg = True
-build_strategy.inner_step = 1
+build_strategy.inner_step = 10
 strategy = build_strategy.create_fl_strategy()
 
 # endpoints will be collected through the cluster

paddle_fl/examples/gru4rec_demo/fl_scheduler.py

+from paddle_fl.core.scheduler.agent_master import FLScheduler
+
+worker_num = 4
+server_num = 1
+scheduler = FLScheduler(worker_num,server_num)
+scheduler.set_sample_worker_num(4)
+scheduler.init_env()
+print("init env done.")
+scheduler.start_fl_training()

paddle_fl/examples/gru4rec_demo/fl_server.py

@@ -21,5 +21,7 @@ server_id = 0
 job_path = "fl_job_config"
 job = FLRunTimeJob()
 job.load_server_job(job_path, server_id)
+job._scheduler_ep = "127.0.0.1:9091"
 server.set_server_job(job)
+server._current_ep = "127.0.0.1:8181"
 server.start()

paddle_fl/examples/gru4rec_demo/fl_trainer.py

@@ -14,7 +14,9 @@ train_file_dir = "mid_data/node4/%d/" % trainer_id
 job_path = "fl_job_config"
 job = FLRunTimeJob()
 job.load_trainer_job(job_path, trainer_id)
+job._scheduler_ep = "127.0.0.1:9091"
 trainer = FLTrainerFactory().create_fl_trainer(job)
+trainer._current_ep = "127.0.0.1:{}".format(9000+trainer_id)
 trainer.start()
 
 r = Gru4rec_Reader()
@@ -25,10 +27,14 @@ step_i = 0
 while not trainer.stop():
     step_i += 1
     print("batch %d start train" % (step_i))
+    train_step = 0
     for data in train_reader():
         #print(np.array(data['src_wordseq']))
         ret_avg_cost = trainer.run(feed=data,
                     fetch=["mean_0.tmp_0"])
+        train_step += 1
+	if train_step == trainer._step:
+		break
         avg_ppl = np.exp(ret_avg_cost[0])
         newest_ppl = np.mean(avg_ppl)
         print("ppl:%.3f" % (newest_ppl))

paddle_fl/examples/gru4rec_demo/run.sh

@@ -2,6 +2,7 @@ unset http_proxy
 unset https_proxy
 python fl_master.py
 sleep 2
+python -u fl_scheduler.py >scheduler.log &
 python -u fl_server.py >server0.log &
 sleep 2
 python -u fl_trainer.py 0 >trainer0.log &

setup.py

@@ -37,6 +37,8 @@ if max_version < 3:
 else:
     REQUIRED_PACKAGES += ["numpy"]
 
+REQUIRED_PACKAGES += ["unittest2"]
+
 setup(
     name='paddle_fl',
     version=fl_version.replace('-', ''),