How do I use TensorBoard for my machine learning projects?

How do I use TensorBoard for my machine learning projects?

TensorBoard is a powerful tool for visualizing metrics in TensorFlow. You can start by importing it in your script.

What kind of metrics can I visualize with TensorBoard?

You can visualize metrics like loss, accuracy, histograms of weights, and even images or text data.

Can you show me a simple code example to get started?

Sure! Here's a basic example:

import tensorflow as tf
from tensorflow import keras

# Define a simple model
model = keras.Sequential([...])

# Compile the model
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])

# Create a TensorBoard callback
tensorboard_callback = keras.callbacks.TensorBoard(log_dir='./logs')

# Train the model with the callback
model.fit(x_train, y_train, epochs=5, callbacks=[tensorboard_callback])

After running this, you can start TensorBoard with tensorboard --logdir=./logs.

How do I access TensorBoard in my browser?

Once you start TensorBoard, it will usually be accessible at http://localhost:6006 in your web browser.