add scheduler in SECAGG and fix compatibility between py2 and py3

docs/requirements.txt

@@ -2,3 +2,4 @@ sphinx==2.1.0
 mistune
 sphinx_rtd_theme
 paddlepaddle>=1.6
+zqm

paddle_fl/core/trainer/diffiehellman/diffiehellman.py

@@ -42,9 +42,9 @@ try:
     from ssl import RAND_bytes
     rng = RAND_bytes
 except(AttributeError, ImportError):
-    raise RNGError
+    rng = os.urandom
+    #raise RNGError
 #python2
-#rng = os.urandom
 
 class DiffieHellman:
     """
@@ -115,13 +115,13 @@ class DiffieHellman:
         self.shared_secret = pow(other_public_key,
                                  self.private_key,
                                  self.prime)
-        #python2
-        #length = self.shared_secret.bit_length() // 8 + 1
-        #shared_secret_as_bytes = ('%%0%dx' % (length << 1) % self.shared_secret).decode('hex')[-length:]
-
-        #python3
-        shared_secret_as_bytes = self.shared_secret.to_bytes(self.shared_secret.bit_length() // 8 + 1, byteorder='big')
-
+        try:
+            #python3
+            shared_secret_as_bytes = self.shared_secret.to_bytes(self.shared_secret.bit_length() // 8 + 1, byteorder='big')
+        except:
+            #python2
+            length = self.shared_secret.bit_length() // 8 + 1
+            shared_secret_as_bytes = ('%%0%dx' % (length << 1) % self.shared_secret).decode('hex')[-length:]
         _h = sha256()
         _h.update(bytes(shared_secret_as_bytes))
 

paddle_fl/core/trainer/fl_trainer.py

@@ -16,7 +16,8 @@ import logging
 from paddle_fl.core.scheduler.agent_master import FLWorkerAgent
 import numpy
 import hmac
-#from .diffiehellman.diffiehellman import DiffieHellman
+import hashlib
+from .diffiehellman.diffiehellman import DiffieHellman
 
 class FLTrainerFactory(object):
     def __init__(self):
@@ -190,6 +191,8 @@ class SecAggTrainer(FLTrainer):
         self._step_id = s
 
     def start(self):
+        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
@@ -219,7 +222,7 @@ class SecAggTrainer(FLTrainer):
             self._logger.debug("begin to run send program")
             noise = 0.0
             scale = pow(10.0, 5)
-            digestmod="SHA256"
+            digestmod=hashlib.sha256
             # 1. load priv key and other's pub key
             dh = DiffieHellman(group=15, key_length=256)
             dh.load_private_key(self._key_dir + str(self._trainer_id) + "_priv_key.txt")
@@ -245,5 +248,3 @@ class SecAggTrainer(FLTrainer):
         self.cur_step += 1
         return loss
 
-    def stop(self):
-        return False

paddle_fl/examples/secagg_demo/fl_scheduler.py

+from paddle_fl.core.scheduler.agent_master import FLScheduler
+
+worker_num = 2
+server_num = 1
+scheduler = FLScheduler(worker_num,server_num)
+scheduler.set_sample_worker_num(worker_num)
+scheduler.init_env()
+print("init env done.")
+scheduler.start_fl_training()

paddle_fl/examples/secagg_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/secagg_demo/fl_trainer.py

@@ -28,8 +28,10 @@ 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.trainer_id = trainer_id
+trainer._current_ep = "127.0.0.1:{}".format(9000+trainer_id)
 trainer.trainer_num = trainer_num
 trainer.key_dir = "./keys/"
 trainer.start()

paddle_fl/examples/secagg_demo/run.sh

@@ -5,10 +5,12 @@ if [ ! -d log ];then
     mkdir log
 fi
 
-python3 fl_master.py
+python fl_master.py
 sleep 2
-python3 -u fl_server.py >log/server0.log &
+python -u fl_server.py >log/server0.log &
 sleep 2
-python3 -u fl_trainer.py 0 >log/trainer0.log &
+python -u fl_scheduler.py > log/scheduler.log &
 sleep 2
-python3 -u fl_trainer.py 1 >log/trainer1.log &
+python -u fl_trainer.py 0 >log/trainer0.log &
+sleep 2
+python -u fl_trainer.py 1 >log/trainer1.log &