# শুধুমাত্র টেন্সরফ্লো ২.x এর জন্য
try:
  # %tensorflow_version only exists in Colab.
  %tensorflow_version 2.x
except Exception:
  pass
​
from __future__ import absolute_import, division, print_function, unicode_literals
​
# টেন্সরফ্লো ২.x এবং tf.keras
import tensorflow as tf
from tensorflow import keras
​
# সাহায্যকারী লাইব্রেরি
import numpy as np
import matplotlib.pyplot as plt
import random
​
print(tf.__version__)

TensorFlow 2.x selected.
2.1.0-rc1

# কেরাস থেকে এপিআই দিয়ে ডাউনলোড করে দেখাচ্ছি, টিএফডিএস নয়, ভ্যারিয়েশন
# ট্রেনিং এবং টেস্ট ডেটাসেট
mnist = keras.datasets.mnist
(train_images, train_labels), (test_images, test_labels) = mnist.load_data()

Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz
11493376/11490434 [==============================] - 0s 0us/step

# আগের মতো নর্মালাইজ করে ০ থেকে ১ এর মধ্যে রাখছি
train_images = train_images / 255.0
test_images = test_images / 255.0
print('Pixels are normalized')

Pixels are normalized

# শুরুতে মডেল আর্কিটেকচার ডিফাইন করছি
model = keras.Sequential([
  keras.layers.InputLayer(input_shape=(28, 28)),
  keras.layers.Reshape(target_shape=(28, 28, 1)),
  keras.layers.Conv2D(filters=12, kernel_size=(3, 3), activation=tf.nn.relu),
  keras.layers.MaxPooling2D(pool_size=(2, 2)),
  keras.layers.Flatten(),
  keras.layers.Dense(10, activation=tf.nn.softmax)
])
​
# মডেলকে ট্রেইন করতে হবে, তার আগে কম্পাইল
model.compile(optimizer='adam',
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy'])
​
# ডিজিট ক্লাসিফিকেশন মডেল ট্রেনিং
model.fit(train_images, train_labels, epochs=5)

Train on 60000 samples
Epoch 1/5
60000/60000 [==============================] - 11s 177us/sample - loss: 0.2655 - accuracy: 0.9264
Epoch 2/5
60000/60000 [==============================] - 4s 67us/sample - loss: 0.1035 - accuracy: 0.9708
Epoch 3/5
60000/60000 [==============================] - 4s 72us/sample - loss: 0.0759 - accuracy: 0.9785
Epoch 4/5
60000/60000 [==============================] - 4s 68us/sample - loss: 0.0626 - accuracy: 0.9818
Epoch 5/5
60000/60000 [==============================] - 4s 68us/sample - loss: 0.0540 - accuracy: 0.9839
​
<tensorflow.python.keras.callbacks.History at 0x7fc3f04e26a0>

model.summary()

Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
reshape (Reshape)            (None, 28, 28, 1)         0         
_________________________________________________________________
conv2d (Conv2D)              (None, 26, 26, 12)        120       
_________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 13, 13, 12)        0         
_________________________________________________________________
flatten (Flatten)            (None, 2028)              0         
_________________________________________________________________
dense (Dense)                (None, 10)                20290     
=================================================================
Total params: 20,410
Trainable params: 20,410
Non-trainable params: 0
_________________________________________________________________

test_loss, test_acc = model.evaluate(test_images, test_labels)
​
print('Test accuracy:', test_acc)

10000/10000 [==============================] - 1s 68us/sample - loss: 0.0647 - accuracy: 0.9801
Test accuracy: 0.9801

# কনভার্টারের ব্যবহার
converter = tf.lite.TFLiteConverter.from_keras_model(model)
tflite_float_model = converter.convert()
​
# মডেলের সাইজ কিলোবাইটে
float_model_size = len(tflite_float_model) / 1024
print('Float model size = %dKBs.' % float_model_size)

Float model size = 81KBs.

# ফাইলটাকে সেভ করে রাখি এখানে mnist.tflite হিসেবে
f = open('mnist.tflite', "wb")
f.write(tflite_float_model)
f.close()
​
# ডাউনলোড করি mnist.tflite হিসেবে
from google.colab import files
files.download('mnist.tflite')
​
print('`mnist.tflite` has been downloaded')

`mnist.tflite` has been downloaded