fabclmnt

import tensorflow as tf

# A simple convolutional neural network model 
model = tf.keras.Sequential()
# A convolutional layer
model.add(tf.keras.layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32, 32, 3)))
# A pooling layer
model.add(tf.keras.layers.MaxPooling2D((2, 2)))
# Flattening the above result
model.add(tf.keras.layers.Flatten())

fabclmnt

import tensorflow as tf
from tensorflow.keras import layers, utils, models

# Input layer
read = layers.Input(shape=(64,64,1))
# First feature extractor
conv1 = layers.Conv2D(16, kernel_size=3, activation='relu')(read)
pool1 = layers.MaxPooling2D(pool_size=(2, 2))(conv1)
flat1 = layers.Flatten()(pool1)
# Second feature extractor

fabclmnt

import tensorflow as tf

class MyModel(tf.keras.Model):
 def __init__(self):
   super(MyModel, self).__init__(name='my_classifier')
   # The layers are defined here
   self.dense_1 = tf.keras.layers.Dense(16)
   self.dense_2 = tf.keras.layers.Dense(1, activation='sigmoid')
 def call(self, inputs):
   # The forward pass is defined here

fabclmnt

model.compile(
  # Mean Squared Error is used as loss function
  loss = tf.keras.losses.MeanSquaredError(),
  # If you do not pass parameters the default ones are used for Adam()
  optimizer = tf.keras.optimizers.Adam(),
  # Metrics are passed as a list seperated by commas.
  metrics = ['accuracy']
)

model.fit(

fabclmnt

callbacks = [
    tf.keras.callbacks.EarlyStopping(
        # Monitoring the accuracy metric
        monitor='accuracy',
        # Specifies that the accuracy should improve by 1e-5. Which means that the absolute differnce
        # between the current and the last value of accuracy should surpass min_delta
        min_delta=1e-5,
        # Training is stopped if accuracy does not imaprove for 5 epochs
        patience=5
        )

fabclmnt

# Directory storing the images placed into folder such that the folder name corresponds
# to the class name inside of the folder
train_dir = '/tmp/data/'

# ImageDataGenerator can be used to augment images by rotation_range, zoom_range. 
#The line below will scale pixel values between [0,1] 
ImageDataGenerator(rescale = 1./255)

# Flow training images in batches of 20
train_generator = datagen.flow_from_directory(

fabclmnt

import tensorflow as tf

# Spread training across multiple GPUs
strategy = tf.distribute.MirroredStrategy()

with strategy.scope():
  model = tf.keras.Sequential([
          # Define model here
          ])
  # Compile model

fabclmnt

import tensorflow as tf

# Create and train your model
model.fit()

# Save model using savedModel. It saves the complete model
tf.saved_model.save(model, "directory/to/model_dir")

fabclmnt

import tensorflow as tf

# Load model
model = tf.saved_model.load("directory/to/model_dir")

fabclmnt

import numpy as np
np.random.seed(0)

import tensorflow as tf
try:
    tf.get_logger().setLevel('INFO')
except Exception as exc:
    print(exc)
import warnings
warnings.simplefilter("ignore")