def on_request_update(*args):
req = args[0]
# req:
# 'model_id'
# 'round_number'
# 'current_weights'
# 'weights_format'
# 'run_validation'
print("update requested")
if req['weights_format'] == 'pickle':
weights = pickle_string_to_obj(req['current_weights'])
self.local_model.set_weights(weights)
my_weights, train_loss, train_accuracy = self.local_model.train_one_round()
resp = {
'round_number': req['round_number'],
'weights': obj_to_pickle_string(my_weights),
'train_size': 10,#self.local_model.x_train.shape[0],
'valid_size': 0,#self.local_model.x_valid.shape[0],
'train_loss': train_loss,
'train_accuracy': train_accuracy,
}
if req['run_validation']:
valid_loss, valid_accuracy = self.local_model.validate()
resp['valid_loss'] = valid_loss
resp['valid_accuracy'] = valid_accuracy
self.sio.emit('client_update', resp)
def on_request_update(*args): # 收到 request_update
req = args[0] # 获得数据 包含当前模型权重
# req:
# 'model_id'
# 'round_number'
# 'current_weights'
# 'weights_format'
# 'run_validation'
print("update requested")
if FederatedClient.SLEEP_MODE:
self.intermittently_sleep(p=1., low=10,
high=100) # randowm sleep
if req['weights_format'] == 'pickle': # 解码 获得模型权重
weights = pickle_string_to_obj(req['current_weights'])
self.local_model.set_weights(weights) # 设定权重
my_weights, train_loss, train_accuracy = self.local_model.train_one_round(
) # 获得局部训练集上的表现
resp = {
'round_number': req['round_number'],
'weights': obj_to_pickle_string(my_weights),
'train_size': self.local_model.x_train.shape[0],
'valid_size': self.local_model.x_valid.shape[0],
'train_loss': train_loss,
'train_accuracy': train_accuracy,
}
if req['run_validation']: # 是否需要测 验证集
valid_loss, valid_accuracy = self.local_model.validate()
resp['valid_loss'] = valid_loss
resp['valid_accuracy'] = valid_accuracy
self.sio.emit('client_update', resp) # 回复 client_update事件 附带所需的数据
def on_stop_and_eval(*args):
req = args[0]
if req['weights_format'] == 'pickle':
weights = pickle_string_to_obj(req['current_weights'])
self.local_model.set_weights(weights)
test_loss, test_accuracy = self.local_model.evaluate()
resp = {
'test_size': self.local_model.x_test.shape[0],
'test_loss': test_loss,
'test_accuracy': test_accuracy
}
self.sio.emit('client_eval', resp)
def on_stop_and_eval(*args): # 收到 stop_and_eval
req = args[0] # 获得数据 包含当前模型权重
if req['weights_format'] == 'pickle': # 解码 获得权重
weights = pickle_string_to_obj(req['current_weights'])
self.local_model.set_weights(weights) # 更新局部模型
test_loss, test_accuracy = self.local_model.evaluate() # 获得局部测试集表现
resp = {
'test_size': self.local_model.x_test.shape[0],
'test_loss': test_loss,
'test_accuracy': test_accuracy
}
self.sio.emit('client_eval',
resp) # 向服务端 发送事件 client_eval 附带数据(局部测试集上的表现)
def on_request_update(*args):
req = args[0]
# req:
# 'model_id'
# 'round_number'
# 'current_weights'
# 'weights_format'
# 'run_validation'
print("update requested")
time1 = time.time()
#if req['weights_format'] == 'pickle':
# weights = pickle_string_to_obj(req['current_weights'])
weights = pickle_string_to_obj(req['current_weights'])
r = req['round_number']
self.local_model.set_weights(weights)
my_weights, train_loss, train_accuracy = self.local_model.train_one_round(
r, self.cid)
resp = {
'round_number': req['round_number'],
'weights': obj_to_pickle_string(my_weights),
'train_size': self.local_model.x_train.shape[0],
'valid_size': self.local_model.x_valid.shape[0],
'train_loss': train_loss,
'train_accuracy': train_accuracy,
}
time2 = time.time()
#print("local training time is ", time2-time1)
# if req['run_validation']:
# valid_loss, valid_accuracy = self.local_model.validate()
# resp['valid_loss'] = valid_loss
# resp['valid_accuracy'] = valid_accuracy
self.sio.emit('client_update', resp)
def on_request_update(*args):
time_start_request_update = time.time()
with open(self.fo_name, 'a') as fo:
fo.write("time_start_request_update: " +
str(time_start_request_update) + "\n")
print(
"------------------------------------------------time_start_request_update: ",
time_start_request_update - self.time_start)
req = args[0]
# req:
# 'model_id'
# 'round_number'
# 'current_weights'
# 'weights_format'
# 'run_validation'
print("update requested")
if req['weights_format'] == 'pickle':
weights = pickle_string_to_obj(req['current_weights'])
self.local_model.set_weights(weights)
time_start_training = time.time()
my_weights, train_loss, train_accuracy = self.local_model.train_one_round(
)
time_end_training = time.time()
with open(self.f_training_name, 'a') as f_training:
f_training.write("time_training: " +
str(time_end_training - time_start_training) +
"\n")
resp = {
'round_number': req['round_number'],
'weights': obj_to_pickle_string(my_weights),
'train_size': self.local_model.x_train.shape[0],
'valid_size': self.local_model.x_valid.shape[0],
'train_loss': train_loss,
'train_accuracy': train_accuracy,
}
if req['run_validation']:
valid_loss, valid_accuracy = self.local_model.validate()
resp['valid_loss'] = valid_loss
resp['valid_accuracy'] = valid_accuracy
time_start_emit = time.time()
print(
"------------------------------------------------time_start_emit: ",
time_start_emit - self.time_start)
with open(self.fo_name, 'a') as fo:
fo.write("time_start_emit: " + str(time_start_emit) + "\n")
self.sio.emit('client_update', resp)
time_finish_emit = time.time()
#fo.write("time_finish_emit: " + str(time_finish_emit) + "\n")
print(
"------------------------------------------------time_finish_emit: ",
time_finish_emit - self.time_start)
