def getLstm(data):
mode = data['mode']
modelType = data['type']
client_id = data['client_id']
new_lstm = glstm.gx_lstm()
if mode == 'train':
gmc = gx_model_ctr.GMC()
gmc.ensure_path(modelType, gx_model_name, client_id)
base_dir = gmc.get_base_dir()
x_train, y_train = dp.getTrainData(data['data_train'])
data['x_train'] = x_train
data['y_train'] = y_train
new_lstm.create_lstm(data)
new_lstm.save_lstm(base_dir)
print('------train success------')
result = 'success'
res = {'res_train': result}
K.clear_session()
if mode == 'predict':
gmc = gx_model_ctr.GMC()
tag, base_dir = gmc.ensure_model(modelType, gx_model_name, client_id)
predictX, scaler, m = dp.getTestdata(data['predictX'])
model = new_lstm.load_lstm(base_dir + 'model')
testPredict = model.predict(predictX)
res_predict = dp.datatransfrom(testPredict, scaler, m)
res = {'res_predict': list(res_predict)}
K.clear_session()
return res
def predict(self, x_data, gx_model_name, modelType, client_id, model):
gmc = gx_model_ctr.GMC()
tag, base_dir = gmc.ensure_model(modelType, gx_model_name, client_id)
ids = gx_model_name + '-' + modelType + '-' + client_id
y_data = []
if ids in model:
print('tag:', "model in Memory")
p_label, p_acc, p_val = svm_predict(y_data, x_data, model[ids][1])
res = {
'p_acc': str(p_acc),
'p_label': str(p_label),
'p_val': str(p_val)
}
result = res
elif tag == False:
result = {'error:': 'please train model !'}
else:
print('tag:', "model in file")
model = svm_load_model(base_dir + 'model')
p_label, p_acc, p_val = svm_predict(y_data, x_data, model)
res = {
'p_acc': str(p_acc),
'p_label': str(p_label),
'p_val': str(p_val)
}
result = res
return result
def train_two(x_data,y_data,train_epoches,modelType,client_id):
gmc=gx_model_ctr.GMC()
gmc.ensure_path(modelType,gx_model_name,client_id)
base_dir=gmc.get_base_dir()
network=gx_network.GX_network()
network.load_network(base_dir)
loss=network.train(x_data,y_data,train_epoches)
network.save_network(base_dir)
ids=gx_model_name+'-'+modelType+'-'+client_id
train_time=time.time()
model[ids]={0:train_time,1:network}
return loss
def train_one(x_data,y_data,learn_rate,layers,train_epoches,modelType,client_id):
gmc=gx_model_ctr.GMC()
gmc.ensure_path(modelType,gx_model_name,client_id)
base_dir=gmc.get_base_dir()
x_data=np.array(x_data)
y_data=np.array(y_data)
structs={'input_ps':x_data[0].size,'learn_rate':learn_rate,'layers':layers}
get_model=getModel.Model()
get_model.save_model(structs,x_data,y_data,train_epoches,base_dir,modelType,client_id,gx_model_name)
train_time,network,loss,ids=get_model.get_model()
model[ids]={0:train_time,1:network}#保存模型
return loss
def prdict(self, x_data, gx_model_name, modelType, client_id, model):
gmc = gx_model_ctr.GMC()
tag, base_dir = gmc.ensure_model(modelType, gx_model_name, client_id)
ids = gx_model_name + '-' + modelType + '-' + client_id
if ids in model:
print('tag:', "model in Memory")
res = model[ids][1].predict(x_data)
result = res.tolist()
elif tag == False:
result = {'error:': 'please train model !'}
else:
print('tag:', "model in file")
network = networks.GX_network()
network.load_network(base_dir + 'model')
res = network.predict(x_data)
result = res.tolist()
return result
def prdict(self, x_data, gx_model_name, modelType, client_id, Model,
params):
gmc = gx_model_ctr.GMC()
tag, base_dir = gmc.ensure_model(modelType, gx_model_name, client_id)
ids = gx_model_name + '-' + modelType + '-' + client_id
if ids in Model:
print('tag:', "model in Memory")
ypred = Model[ids][1].predict(xgb.DMatrix(x_data))
ypred = np.array(ypred).astype(np.str)
result = {"code": 200, "result": list(ypred)}
elif tag == False:
result = {'error:': 'please train model !'}
else:
print('tag:', "model in file")
model = xgb.Booster(params)
model.load_model(base_dir + 'model')
ypred = model.predict(xgb.DMatrix(x_data))
ypred = np.array(ypred).astype(np.str)
result = {"code": 200, "result": list(ypred)}
return result
"""
Created on Mon Nov 19 16:16:24 2018
@author: Zcy
"""
import os, sys
sys.path = [os.path.dirname(os.path.abspath(__file__))] + sys.path
import gx_svmModel
import util.gx_model_ctr as gx_model_ctr
import schedule
import time
import threading
gx_model_name = 'svm'
gmc = gx_model_ctr.GMC()
gsm = gx_svmModel.Model()
model = {}
def train(data):
gmc.ensure_path(data['type'], gx_model_name, data['client_id'])
base_dir = gmc.get_base_dir()
trainx, trainy = trans2libsvmX(data['train_data'])
options = data['options']
try:
gsm.save_model(trainx, trainy, options, base_dir, data['type'],
data['client_id'], gx_model_name)
train_time, svmModel, ids = gsm.get_model()
model[ids] = {0: train_time, 1: svmModel}
res = {'train': str('success')}