There are three main types of scripts in this repository i.e. scripts for a Pytorch Google Compute Instance (Pytorch folder), scripts for a Tensorflow Google Compute Instance (Tensorflow folder) and a script for setting up essential packages on your local Ubuntu OS (Setup.sh).
In the Tensorflow folder, there are 2 .sh files.
The first file i.e. tf-GCP.sh can be used to create a tensorflow GCP instance with an Nvidia Tesla K-80 GPU, 200GB storage, 4 vCPUs, and 26 GB RAM. It will auto-install the Nvidia driver.
Usage: sh tf-GCP.sh <INSTANCE_NAME>
The second file i.e. tf-ssh.sh can be used to start the Tensorflow instance and ssh into it from the terminal without going to the Google Cloud webpage.
Usage: sh tf-ssh.sh
In the Pytorch folder, there are 2 .sh files.
The first file i.e. pytorch-GCP.sh can be used to create a Pytorch GCP instance with an Nvidia Tesla K-80 GPU, 200GB storage, 4 vCPUs, and 26 GB RAM. It will auto-install the Nvidia driver.
Usage: sh pytorch-GCP.sh <INSTANCE_NAME>
The second file i.e. pyt-ssh.sh can be used to start the Pytorch instance and ssh into it from the terminal without going to the Google Cloud webpage.
Usage: sh tf-ssh.sh
Int the "fastai-local" folder, there is a .sh file that creates and activates a conda fastai environment using the fastai course v3 yml file.
Usage: sh fastai-local.sh