keras_quickstart.ipynb

Kernel: Python 3

Keras quickstart

We recommend running this example in Colab's GPU runtime. It will run on Jax, TensorFlow or PyTorch, simply change the line below.

import os
os.environ["KERAS_BACKEND"] = "jax"  # Or "torch" or "tensorflow"

Train an MNIST classifier with a mini ResNet model

import keras
from keras.datasets import mnist
from keras import layers

(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255., x_test / 255.

inputs = keras.Input(shape=(28, 28, 1))
x = layers.Conv2D(32, 3, activation="relu")(inputs)
x = layers.Conv2D(64, 3, activation="relu")(x)
residual = x = layers.MaxPooling2D(3)(x)

x = layers.Conv2D(64, 3, activation="relu", padding="same")(x)
x = layers.Conv2D(64, 3, activation="relu", padding="same")(x)
x = x + residual

x = layers.Conv2D(64, 3, activation="relu")(x)
x = layers.GlobalAveragePooling2D()(x)
x = layers.Dense(256, activation="relu")(x)
x = layers.Dropout(0.5)(x)
outputs = layers.Dense(10, activation="softmax")(x)

model = keras.Model(inputs, outputs, name="mini_resnet")

keras.utils.plot_model(model, "mini_resnet.png", dpi=100, show_shapes=True)

model.compile(
    optimizer="adam",
    loss="sparse_categorical_crossentropy",
    metrics=["accuracy"],
)
model.fit(x_train, y_train, epochs=20)
model.evaluate(x_test, y_test)

Epoch 1/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 7s 3ms/step - accuracy: 0.6881 - loss: 0.8581
Epoch 2/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 5s 2ms/step - accuracy: 0.9752 - loss: 0.0845
Epoch 3/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9823 - loss: 0.0599
Epoch 4/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9872 - loss: 0.0437
Epoch 5/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9904 - loss: 0.0337
Epoch 6/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9912 - loss: 0.0296
Epoch 7/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9931 - loss: 0.0241
Epoch 8/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9942 - loss: 0.0203
Epoch 9/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9947 - loss: 0.0189
Epoch 10/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9960 - loss: 0.0141
Epoch 11/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9965 - loss: 0.0118
Epoch 12/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9960 - loss: 0.0124
Epoch 13/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9965 - loss: 0.0117
Epoch 14/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9964 - loss: 0.0110
Epoch 15/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9969 - loss: 0.0100
Epoch 16/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9973 - loss: 0.0088
Epoch 17/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9974 - loss: 0.0101
Epoch 18/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9981 - loss: 0.0058
Epoch 19/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9974 - loss: 0.0081
Epoch 20/20
1875/1875 ━━━━━━━━━━━━━━━━━━━━ 4s 2ms/step - accuracy: 0.9978 - loss: 0.0077
313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step - accuracy: 0.9921 - loss: 0.0478

[0.03639788180589676, 0.9934000372886658]