CoCalc Tutorials

About CoCalc Tutorials

Welcome to the CoCalc tutorial page! We will regularly post content regarding various uses of CoCalc’s many features including our Jupyter notebook, LaTeX editor, online Linux environment, and recently created collaborative Whiteboard.

We will provide the basics required to “get started” with CoCalc as an individual user, as well as information beneficial for instructors and students to effectively take advantage of our Course Management System, provide automated feedback and grading via nbgrader, and utilize other tools essential to academically oriented usage.

You may also expect to see an assortment of technical examples using Python, SageMath, R, and Julia within our collaborative notebook interface.

Subscribe to Our Blog and Tutorial Series

Get updates on our newest content delivered directly to your inbox by adding your email below.

Using LLM’s with Ollama with Ollama-webui and a GPU on cocalc.com

A tutorial showing you exactly how to use Ollama and the Ollama-webui to download and run some open large language models. No installation is required and the total compute cost to run the entire demo yourself is only a nickel!

https://cocalc.com/

https://ollama.ai/

https://github.com/ollama-webui/ollam…

Many more details here: https://github.com/sagemathinc/cocalc…

Using PyTorch with a GPU in a Jupyter Notebook Online Via CoCalc

In this short video, William Stein demos how to use CoCalc’s new GPU functionality to run a Jupyter notebook with a GPU and the PyTorch kernel online for one penny!

So, What is CoCalc?

A quick three-minute overview reviewing CoCalc’s general usage and licensing.

Using the Google Colab Environment with and without a GPU Online Via CoCalc

In CoCalc you can use the same Python/R environments that are part of Colab, both with and without a GPU.

https://github.com/sagemathinc/cocalc…

To use Colab with a GPU on CoCalc, create a compute server with a GPU, and select the “Google Colab” image. Once your compute server is running you can open any Jupyter notebook or Terminal, and use the server selector in the upper left to run it on your compute server.

You can then use Colab with GPU acceleration. CoCalc also lets you directly run the Colab environment (and Tensorflow and a million other things!) completely for free without a GPU by just making a Jupyter notebook and NOT running it on a compute server.

This is useful because:

It is a single Python environment with excellent support for both PyTorch and Tensorflow at the same time. It makes it easier for you move code back and forth between Colab and CoCalc, depending on who you are collaborating with.

This ensures the code you write on CoCalc will work on Colab when you share it publicly. You may be able to find better deals on GPU’s in CoCalc than you would directly on hosted Colab.

Note that both Colab and Cocalc have a way to run Jupyter kernels “on prem” on your own server. For CoCalc the steps are exactly as above with a GPU, but instead of choosing a compute server on Google cloud, you select “on prem”, and paste a line of code into a Virtual Machine that you start running on your own server.

Jupyter Widgets in CoCalc

William Stein discusses how Jupyter Widgets are implemented in Cocalc (https://cocalc.com/), including real-time collaboration. Presentation is at https://cocalc.com/wstein/support/ipw…, and other resources are at https://cocalc.com/wstein/support/ipw…. Presentation was given in the 17 May 2022 Jupyter Widgets development meeting.