def generate_seed_model(model_dimension='VGG11'):
model = VGG(dimension=model_dimension)
outfile_name = 'seed/' + model_dimension + '_keras.npz'
weights = model.get_weights()
helper = KerasHelper()
if not os.path.exists('seed'):
os.makedirs('seed')
helper.save_model(weights, outfile_name)
def get_helper(helper_type):
""" Return an instance of the helper class.
:param helper_type (str): The helper type ('keras','pytorch')
:return:
"""
if helper_type == 'numpyarray':
from fedn.utils.numpyarrayhelper import NumpyArrayHelper
return NumpyArrayHelper()
elif helper_type == 'keras':
from fedn.utils.kerashelper import KerasHelper
return KerasHelper()
elif helper_type == 'pytorch':
from fedn.utils.pytorchhelper import PytorchHelper
return PytorchHelper()
else:
return None
def train(model, filename, settings):
print("-- RUNNING TRAINING --", flush=True)
train_x, _, _ = read_data(filename)
model.fit(train_x,
train_x,
batch_size=settings['batch_size'],
epochs=settings['epochs'])
return model
if __name__ == '__main__':
with open('settings.yaml', 'r') as fh:
try:
settings = dict(yaml.safe_load(fh))
except yaml.YAMLError as e:
raise (e)
from fedn.utils.kerashelper import KerasHelper
from models.autocoder import create_seed_model
helper = KerasHelper()
weights = helper.load_model(sys.argv[1])
model = create_seed_model()
model.set_weights(weights)
model = train(model, '../data/train.csv', settings)
helper.save_model(model.get_weights(), sys.argv[2])
from fedn.utils.kerashelper import KerasHelper
from models.AMLmodel import construct_model
if __name__ == '__main__':
# Create a seed model and push to Minio
model = construct_model()
outfile_name = "initial_model.npz"
weights = model.get_weights()
helper = KerasHelper()
helper.save_model(weights, outfile_name)
if __name__ == '__main__':
with open('settings.yaml', 'r') as fh:
try:
settings = dict(yaml.safe_load(fh))
except yaml.YAMLError as e:
raise (e)
with open('/app/client_settings.yaml', 'r') as fh:
try:
client_settings = dict(yaml.safe_load(fh))
except yaml.YAMLError as e:
raise (e)
from fedn.utils.kerashelper import KerasHelper
helper = KerasHelper()
weights = helper.load_model(sys.argv[1])
model = VGG(dimension=settings['model_dimension'])
opt = keras.optimizers.Adam(learning_rate=0.001)
model.compile(loss='categorical_crossentropy',
optimizer=opt,
metrics=['accuracy'])
model.set_weights(weights)
report = validate(model, '/app/data', settings)
with open(sys.argv[2], "w") as fh:
fh.write(json.dumps(report))
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
# File : init_model.py
# Author : Sheetal Reddy <*****@*****.**>
# Date : 08.03.2021
# Last Modified Date: 08.03.2021
# Last Modified By : Sheetal Reddy <*****@*****.**>
import sys
import os
from client1_new import TrainingProcess, Model, TrainDataReader
from fedn.utils.kerashelper import KerasHelper
from kerasmodel import create_seed_model
if __name__ == '__main__':
outfile_name = '/seed/' + sys.argv[1]
helper = KerasHelper()
start_process = create_seed_model('/client/', pretrained=True)
helper.save_model(start_process.local_model.get_weights(),
path=outfile_name)
print("seed model saved as: ", outfile_name)