training_repo = 'http://deepdetect.com/dd/datasets/mnist_csv/mnist_test.csv'
# service creation
model = {'repository':model_repo}
parameters_input = {'connector':'csv'}
parameters_mllib = {}
parameters_output = {}
dd.put_service(sname,model,description,mllib,
parameters_input,parameters_mllib,parameters_output,'unsupervised')
# training
train_data = [training_repo]
parameters_input = {'id':'','separator':',','label':'label'}
parameters_mllib = {'iterations':500}
parameters_output = {}
predout = dd.post_train(sname,train_data,parameters_input,parameters_mllib,parameters_output,async=True)
time.sleep(1)
train_status = ''
while True:
train_status = dd.get_train(sname,job=1,timeout=3)
if train_status['head']['status'] == 'running':
print train_status['body']['measure']
else:
print train_status
predout = train_status
break
predictions = predout['body']['predictions']
N = len(predictions)
points = np.empty((N,2),dtype=np.float)
parameters_input_service = {'connector':'txt'}
if template == "mlp":
parameters_mllib_service = {'template':template,'nclasses':nclasses,'layers':layers,'activation':activation}
elif template == "lregression":
parameters_mllib_service = {'template':template,'nclasses':nclasses,'activation':activation}
parameters_output_service = {'measure':['mcll','f1']}
dd.put_service(service_name,model,description,mllib,parameters_input_service,parameters_mllib_service,parameters_output_service)
#Start training the service
iterations = int(service['iterations'])
solver_type = service['solver_type']
base_lr = float(service['base_lr'])
parameters_input_training = {'shuffle':True,'test_split':test_split,'min_count':min_count,'min_word_length':min_word_length,'count':False}
parameters_mllib_training = {'gpu':True,'solver':{'iterations':iterations,'test_interval':test_interval,'base_lr':base_lr,'solver_type':solver_type},'net':{'batch_size':batch_size}}
parameters_output_training = {'measure':['mcll','f1','cmdiag','cmfull']}
train_data = [root_repository+'dataset/']
training_service = dd.post_train(service_name.lower(),train_data,parameters_input_training,parameters_mllib_training,parameters_output_training,async=True)
job_number = training_service['head']['job']
#Get training data while the service is running
sleep(20)
status_code = 200
count_job_data = 1
while status_code == 200:
job_data = dd.get_train(service_name.lower(),job=job_number, measure_hist=True)
status_code = job_data['status']['code']
if not 'accp' in job_data['body']['measure']:
sleep(20)
continue
if job_data['head']['status'] == 'running':
log_file.write("job running time "+str(job_data['head']['time'])+"\n")
log_file.write("Iteration number "+str(job_data['body']['measure']['iteration'])+"\n")
log_file.flush()
