gsgbills/deeplearning_keras2

Modification of fast.ai deep learning course notebooks for usage with Keras 2 and Python 3.

Modified notebooks and Python files for Keras 2 and Python 3 from the fast.ai Deep Learning course

The repository includes modified copies of the original Jupyter notebooks and Python files from the excellent (and really unique) deep learning course "Practical Deep Learning For Coders" Part 1 and Part 2, created by fast.ai. The original files require Keras 1.

Part 1

Located in the nbs folder. Tested with Keras 2.0.6 on both Ubuntu 16.04 and Python 3.5 (installed through apt-get) and MacOS 10.12.4 with Python 3.6 (installed with Homebrew). In Part 1 the Theano backend for Keras has been used.

Part 2

Located in the nbs2 folder. Tested with Keras 2.0.6 on Ubuntu 16.04 with Python 3.5 (installed through apt-get). In Part 2 the TensorFlow backend for Keras has been used.

The files keras.json.for_TensorFlow and keras.json.for_Theano provide a template for the appropriate keras.json file, based on which one of the two backends needs to be used.

A Python 3 virtualenv has been used for both parts. In order to facilitate the installation of the required Python packages, this repository includes also the requirement files that can be used with the pip command. These files include additional packages that might be useful for further exploration.

The comments that I inserted in the modules generally start with "# -" when they are not just "# Keras 2".

Notes about Part 2

Issues

rossman.ipynb: section "Using 3rd place data" has been left out for lack of the required data

spelling_bee_RNN.ipynb: after the main part of the notebook, in the final "Test code ..." section I was not able to solve an issue with the K.conv1d cell not working

taxi_data_prep_and_mlp.ipynb: section "Uh oh ..." has been left out. Caveat: running all the notebook at once exhausted 128 GB RAM; I was able to run each section individually only after resetting the notebook kernel each time

tiramisu-keras.ipynb: in order to run the larger size model I had to reset the notebook kernel in order to free up enough GPU memory (almost 12 GB) and jump directly to the model

Left-out modules

neural-style-pytorch.ipynb (found no way to load the VGG weights; it looks like some version compatibility issue)

rossman_exp.py

seq2seq-translation.ipynb

taxi.ipynb

tiramisu-pytorch.ipynb