Tips on TensorFlow

Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!

Installation¶

The article Stop Installing Tensorflow using pip for performance sake! suggest installing TensorFlow using conda instead of pip as the version installed by conda leverages Intel Math Kernel Library and is about 8 times faster on CPU.

https://www.anaconda.com/tensorflow-in-anaconda/

cuDNN is required by TensorFlow

:::bash
pip install tensorflow

https://www.tensorflow.org/install/gpu

https://www.tensorflow.org/install/docker

https://www.tensorflow.org/install/pip

Docker Images¶

https://github.com/tensorflow/tensorflow/tree/master/tensorflow/tools/dockerfiles/dockerfiles

tensorflow/tools/dockerfiles/dockerfiles/gpu.Dockerfile uses pip (instead of conda) to install Python packages.

General Tips¶

With TensorFlow 2.0, you should use tf.keras instead of the separate Keras package.

https://www.pyimagesearch.com/2019/10/21/keras-vs-tf-keras-whats-the-difference-in-tensorflow-2-0/

Tutorials¶

https://www.tensorflow.org/guide/keras/

https://www.tensorflow.org/tutorials/keras/classification https://www.tensorflow.org/tutorials/quickstart/beginner https://www.tensorflow.org/guide/keras/functional https://www.tensorflow.org/guide/keras/train_and_evaluate https://www.tensorflow.org/guide/keras/custom_layers_and_models https://www.tensorflow.org/guide/keras/masking_and_padding

Inside TensorFlow: tf.Keras (part 1) Inside TensorFlow: tf.Keras (part 2) TensorFlow 2.0 Full Tutorial - Python Neural Networks for Beginners

GPU Runs Out of Memory¶

https://stackoverflow.com/questions/36927607/how-can-i-solve-ran-out-of-gpu-memory-in-tensorflow

https://stackoverflow.com/questions/34199233/how-to-prevent-tensorflow-from-allocating-the-totality-of-a-gpu-memory

https://superuser.com/questions/980216/what-happens-when-the-gpu-memory-is-not-enough

Deep Learning Libraries Based on TensorFlow¶

trax