Jarvislabs.ai

At the time of the writing (3/10/2023) there are only servers in India, and my experience was that the response time in US is slow. Beside that immpediement, it’s an ok service. I still very much prefer to use Paperspace.

Setup

How to set up an environment link. Make sure to add a prefix otherwise environment will be deleted after pausing.

conda create --prefix /home/some_env_name python=3.10 ipykernel mamba -y
conda activate /home/some_env_name

Set up a jupyter kernel:

python -m ipykernel install --user --name=some_env_name

Data management

See https://jarvislabs.ai/docs/data-management/

This is the simplest method just drag and drop one file. Currently folders and not supported.

Using google drive and curlwget: CurlWget.

Set up SSH

Following VSCode tutorial:

  1. First we locally create a private/public SSH key, this can be done using ssh-keygen -t rsa -b 4096 for Jarvis we’ll use former per example

  2. Show the ssh public key: cat ~/.ssh/id_rsa.pub

  3. Copy paste on a Jarvis here

  4. To get the ssh link right click on the instance and copy paste ssh code (this was not obvious): something like ssh -p some_port_name root@ssha.jarvislabs.ai In general we want to use ssh user@hostname to check the connection.