hadifar

model = MyModel()
checkpoint = tf.train.Checkpoint(myModel=model)
checkpoint.save('./test/model.ckpt')
del model

model = MyModel()
checkpoint = tf.train.Checkpoint(myModel=model)
checkpoint.restore(tf.train.latest_checkpoint('./test/'))

hadifar

app = Flask(__name__)
cors = CORS(app)


@app.route("/api/predict", methods=['POST'])
def predict():
    start = time.time()

    data = request.data.decode("utf-8")
    if data == "":

hadifar

DEF_SERVING_DEF_KEY = tf.saved_model.signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY
tensor_x = model.signature_def[DEF_SERVING_DEF_KEY].inputs['x'].name # return Placeholder:0
tensor_y = model.signature_def[DEF_SERVING_DEF_KEY].outputs['y'].name # return 'dense/BiasAdd:0'
x = tf.get_default_graph().get_tensor_by_name(tensor_x)
y = tf.get_default_graph().get_tensor_by_name(tensor_y)

hadifar

import tensorflow as tf

session = tf.Session()

for item in tf.get_default_graph().get_operations():
    print(item)
print(50*'-')
model = tf.saved_model.loader.load(sess=session,
                                   tags=[tf.saved_model.tag_constants.SERVING],
                                   export_dir='./test/')

hadifar

import tensorflow as tf

session = tf.Session()

for item in tf.get_default_graph().get_operations():
    print(item)
print(50*'-')
model = tf.saved_model.loader.load(sess=session,
                                   tags=[tf.saved_model.tag_constants.SERVING],
                                   export_dir='./test/')

hadifar

model = tf.saved_model.loader.load(sess=session,
                                   tags=[tf.saved_model.tag_constants.SERVING],
                                   export_dir='./test/')

hadifar

import numpy as np
import tensorflow as tf

# create dummy data
X_raw = np.array([2013, 2014, 2015, 2016, 2017], dtype=np.float32)
y_raw = np.array([12000, 14000, 15000, 16500, 17500], dtype=np.float32)
X_data = (X_raw - X_raw.min()) / (X_raw.max() - X_raw.min())
Y_data = (y_raw - y_raw.min()) / (y_raw.max() - y_raw.min())
X_data = np.expand_dims(X_data, axis=1)
Y_data = np.expand_dims(Y_data, axis=1)

hadifar

from keras.models import load_model

model.save('my_model.h5')  
del model  # deletes the existing model

# returns a compiled model
# identical to the previous one
model = load_model('my_model.h5')

hadifar

checkpoint = ModelCheckpoint('./test/, monitor='val_acc')
model.fit(x_train, y_train,
          callbacks=[checkpoint])

hadifar

with tf.Session() as sess:
  tf.saved_model.loader.load(sess, [tag_constants.TRAINING], './test/')