TensorFlow & Keras Cheatsheet

1. Basic Imports

import tensorflow as tf


from tensorflow import keras
from keras import layers

2. Build a Sequential Model

model = keras.Sequential([
layers.Dense(128, activation="relu"),
layers.Dense(10, activation="softmax")


])

3. Compile Model

model.compile(
optimizer="adam",
loss="sparse_categorical_crossentropy",
metrics=["accuracy"]


)

4. Train Model

history = model.fit(
x_train, y_train,
batch_size=32,
epochs=10,
validation_split=0.2


)

5. Evaluate

model.evaluate(x_test, y_test)

6. Predict

pred = model.predict(x_sample)

7. Save and Load Models

model.save("model.h5")


restored = keras.models.load_model("model.h5")

8. Custom Layer

class MyLayer(layers.Layer):
def __init__(self):
    super().__init__()

def call(self, inputs):
    return inputs * 2

9. Add Callbacks

checkpoint = keras.callbacks.ModelCheckpoint(
"best.h5", save_best_only=True


)

model.fit(x, y, callbacks=[checkpoint])

10. TensorFlow Dataset Pipeline

ds = tf.data.Dataset.from_tensor_slices((x, y))


ds = ds.shuffle(1000).batch(32).prefetch(tf.data.AUTOTUNE)

11. GPU Check

print(tf.config.list_physical_devices("GPU"))

12. Common Layers

# Dense layer


layers.Dense(64, activation="relu")

Convolution

layers.Conv2D(32, (3,3), activation="relu")

Pooling

layers.MaxPooling2D((2,2))

Dropout

layers.Dropout(0.5)

13. Optimizers

keras.optimizers.Adam()


keras.optimizers.SGD(learning_rate=0.01, momentum=0.9)
keras.optimizers.RMSprop()

14. Loss Functions

keras.losses.MSE


keras.losses.BinaryCrossentropy()
keras.losses.CategoricalCrossentropy()

15. Functional API Example

inputs = keras.Input(shape=(784,))


x = layers.Dense(128, activation="relu")(inputs)
x = layers.Dropout(0.3)(x)
outputs = layers.Dense(10, activation="softmax")(x)

model = keras.Model(inputs, outputs)

16. Transfer Learning

base = keras.applications.ResNet50(
include_top=False,
weights="imagenet",
input_shape=(224,224,3)


)

base.trainable = False # Freeze backbone

model = keras.Sequential([
base,
layers.Flatten(),
layers.Dense(64, activation="relu"),
layers.Dense(10, activation="softmax")
])

17. Export SavedModel

model.save("exported_model")