Jupyter notebooks for demonstration, exploration and prototyping of Effective Quadratures. These can be run on the cloud with Google Colaboratory or Azure Notebooks.
The choice of Google Colab or Azure notebooks us yours. Previewing and sharing notebooks is arguably easier with Colab, but reading/writing persistant data files is arguably easier with Azure. It may come down to whether you have a Microsoft or Google account!
-
First Method - To open notebooks from Github. Go to https://colab.research.google.com switch to Github tab and enter an organization or URL.
-
Second Method - Github jupyter notebooks can directly be accessed on Google Colaboratory by using CPU/GPU/TPU runtimes by replacing https://github.com in the URL by https://colab.research.google.com/github/ e.g. https://colab.research.google.com/github/Effective-Quadratures/EQ-live/blob/master/Blog_posts/computing_moments.ipynb
-
Third Method - A much easier way is to use “Open in Colab” Extension for Chrome.
Edited from https://www.google-colab.com/google-colab-tips-and-tricks/.
Notebooks can be saved to your Google drive. The share button in the top right of the colab interface can be used to share notebooks once saveed. Notebooks can also be pushed to github Gist's or Repo's for sharing.
This is more involved, as persistant storage on the Google Virtual Machine (VM) instance isn't available. Data must be stored on Google drive, which can be mounted within notebooks:
from google.colab import drive
drive.mount('/content/drive',force_remount=True)
%cd "/content/drive/My Drive/<folder name>" ### Change directory to preferred working space
Then temporarily copied to the VM for faster read/write speeds:
!cp -r "/content/drive/My Drive/<folder name>" ./img
For loading data files required in notebooks in this Repo, we instead download them from github with wget
to avoid having to maintain an additional Google drive repository. For example:
if IN_COLAB:
!wget https://raw.githubusercontent.com/Effective-Quadratures/EQ-live/master/Blog_posts/lossarray_sst.npy -O lossarray_sst.npy
Yp = np.load('lossarray_sst.npy').reshape(-1, 1)
The notebooks in this repo can be run on Microsoft Azure notebooks by going here and cloning the Azure notebooks project. Click on the Jupyter notebook you wish to run, select "Python 3.0" or "Python 3.6" as the kernel, and Azure will then load the neccesary dependencies from the requirements.txt file.
Note 1: The notebooks won't be formatted properly if opened in HTML preview mode. You must sign in and run them properly to see figure etc.
Note 2: Initialising the kernel and loading the dependencies (such as the equadratures
and matplotlib
python packages) may take some time, especially when using "Free Compute" resources. Keep an eye on the grey dot in the top right corner, when it goes clear the kernel is "Idle" and you're good to go. For this reason it is recommended that you do not restart the kernel (or rerun the package import cell) when interacting notebooks unless need to.
The Azure notebook project you will clone is itself a clone of this github repo. All the data files stored on this github repo should be accesible from your cloned Azure project. To save data you can download it to a local machine, or push your cloned project to your own github fork of this repo.
There is an "Open terminal" button on Azure notebooks to open a terminal. Standard git commands can be used here to push to your own fork of this github Repo.