ModelBuilt.py

#%%
import tensorflow as tf
from tensorflow.keras.layers import Dense, Input
# import seaborn as sns
import matplotlib.pyplot as plt
import numpy as np
import time
import os
# %matplotlib inline

logdir = os.path.split(os.path.realpath(__file__))[0]
# %%
# generate some data 2-dimension. shape = (10, 2)

def sigmoid(x):
    return 1/(1 + np.exp(-x))

noise = np.random.randn(10, 1)
# print(noise.shape)
W = np.array([[5, 3]]).T
x = np.array(np.arange(10))[None, :].T
# print(x.shape)
x_ = np.ones((10, 1))
X = np.concatenate((x_, x), axis = 1)
X = np.array(X, dtype=np.float32)

# print(X.shape, W.shape)
y = sigmoid(X @ W + noise)
y = np.array(y, dtype=np.float32)
# print(np.array(y, dtype=np.float32))

# plt.scatter(x, y, s = 10, marker='o', c = 'red')
# plt.show()

data = tf.data.Dataset.from_tensor_slices((X, y))

def parse(x, y):
    
    # change the data type of variables
    x = tf.py_function(lambda z: z, [x], tf.float32)
    y = tf.py_function(lambda z: z, [y], tf.float32)

    # set shape on the dataset info.
    x.set_shape((2,))
    y.set_shape((1,))
    return (x, y)

data = data.map(parse)
data = data.shuffle(10)
data = data.repeat()
data = data.batch(10)

InputLayer = Input(shape=(2, ))
OutputLayer = Dense(1, activation='sigmoid')(InputLayer)
model = tf.keras.Model(inputs=InputLayer, outputs=OutputLayer)

model.compile('adam', loss = [tf.losses.Huber()], metrics=[tf.metrics.MeanSquaredError()])
# model.build(input_shape=(None, 2))
model.fit(data, epochs = 500, steps_per_epoch = 20)

# save second version of model.
savedir = './save/Toy/3'
tf.saved_model.save(model, savedir)

# check if saved model is correct.
new_model = tf.saved_model.load(savedir)
# print(list(new_model.signatures.keys()))
# print(new_model.summary())
# infer = new_model.signatures["serving_default"]
# infer(list or dict or array)
print(new_model(tf.constant([[1., 1.]])))