Repo to store my code for the cDiscount Kaggle Competition. Process: - The data from Kaggle comes in .bson files. Run store_data_as_images.py to save image files. - Run images_to_bottlenecks.sh to convert the image files into "bottlenecks" (the last layer of the AlexNet deep net for image recognition). - Run create_chunks.py to convert the bottleneck files into 10 files with all the bottleneck data (for faster loading when running TensorFlow). - Run distributed/train_on_gcp.sh to run the final training on GCP. The compute engines need to be running, as per this tutorial: https://cloud.google.com/solutions/running-distributed-tensorflow-on-compute-engine