This repository contains a commandline program for training co-teaching models ((link to paper)[https://papers.nips.cc/paper/8072-co-teaching-robust-training-of-deep-neural-networks-with-extremely-noisy-labels])
data.py
contains method to obtain specific dataset.model.py
contains different neural network models.trainer.py
contains generic trainers for managing the training processestask.py
the command-line program entry-point that accepts multiple arguments and invokes the training procedures defined in the aforementioned files
gcloud ai-platform jobs submit training <job name> --package-path cot --module-name cot.task --region us-east1 --python-version 3.5 --runtime-version 1.14 --job-dir gs://<your-bucket-name> --scale-tier BASIC_GPU