Overview
This guide will show our users how to use the TensorFlow Profiler to profile the execution of your TensorFlow code.
Code example
Copy and paste the following code into tf-profile.py.
from datetime import datetime import os import tensorflow from tensorflow.keras.datasets import mnist from tensorflow import keras from tensorflow.keras import layers (train_images, train_labels), (test_images, test_labels) = mnist.load_data() model = keras.Sequential([ layers.Dense(512, activation="relu"), layers.Dense(10, activation="softmax") ]) model.compile(optimizer="rmsprop", loss="sparse_categorical_crossentropy", metrics=["accuracy"]) train_images = train_images.reshape((60000, 28 * 28)) train_images = train_images.astype("float32") / 255 test_images = test_images.reshape((10000, 28 * 28)) test_images = test_images.astype("float32") / 255 # Create a TensorBoard callback logs = "logs/" + datetime.now().strftime("%Y%m%d-%H%M%S") tboard_callback = tensorflow.keras.callbacks.TensorBoard(log_dir = logs, histogram_freq = 1, profile_batch = '10,20') model.fit(train_images, train_labels, epochs=10, batch_size=128, callbacks = [tboard_callback])
The tensorflow.keras.callbacks.TensorBoard command will create a tensorboard callback and profile_batch will pick batch number 10 to batch number 20.
Local profiling on your own computer
Run the code with command
python tf-profile.py
Compress the logs folder
tar -zcvf ./logs.tar.gz ./logs
Download the tarball file with sftp and/or hal-ondemand.
Decompress the tarball file
tar -zxvf ./logs.tar.gz
Install the tensorboard profile plugin in your python environment.
pip install tensorboard_plugin_profile
Launch the tensorboard with profiler installed.
tensorboard --logdir ./logs
Remote Profiling on HAL system
Coming soon...