TensorFlow Basics

Introduction

TensorFlow is a comprehensive platform for building and deploying machine learning models with efficient computation graphs.

Tensors

import tensorflow as tf

# Constant tensor
scalar = tf.constant(5)
vector = tf.constant([1, 2, 3])
matrix = tf.constant([[1, 2], [3, 4]])

# Variable (trainable)
var = tf.Variable(tf.random.normal([10, 10]))

# Data types
float32_tensor = tf.constant([1.0], dtype=tf.float32)
int32_tensor = tf.constant([1], dtype=tf.int32)

Eager Execution

# Eager execution is default in TF2
a = tf.constant([[1, 2], [3, 4]])
b = tf.constant([[5, 6], [7, 8]])

result = tf.matmul(a, b)
print(result)  # Immediate evaluation

tf.function

@tf.function
def matmul(a, b):
    return tf.matmul(a, b)

# Compiled to graph for faster execution
result = matmul(a, b)

Autograd

x = tf.Variable(3.0)

with tf.GradientTape() as tape:
    y = x**2

gradient = tape.gradient(y, x)
print(gradient)  # tf.Tensor(6.0, shape=(), dtype=float32)

# For trainable variables
w = tf.Variable(tf.random.normal([10, 5]))
b = tf.Variable(tf.zeros([5]))

with tf.GradientTape() as tape:
    y = tf.matmul(x, w) + b
    loss = tf.reduce_mean(y**2)

grads = tape.gradient(loss, [w, b])

GPU Acceleration

# Check GPU availability
print(tf.config.list_physical_devices('GPU'))

# Move tensor to GPU
if tf.config.list_physical_devices('GPU'):
    with tf.device('/GPU:0'):
        a = tf.constant([[1, 2], [3, 4]])
        result = a * 2

Practice Problems

1. Create tensors of different shapes
2. Use tf.function for optimization
3. Compute gradients
4. Move tensors to GPU
5. Use TensorFlow Hub