Add Support for common ML and Data analysis workloads
Development around Machine Learning and/or tools for performing ML should also be considered well, specifically around building tools for ML which do not directly involve training (such as experiment tracking, data analysis, etc.)
Suvaditya Mukherjee
shared this idea
This is a great suggestion! We are going to investigate what support here can look like and follow-up with you further for more details.
-
Amir F commented
I would love a dev experience like this:
1. clone a git repo to start the project
2. pick the accelerator (eg. GPU, TPU, etc. similar to Colab) so the instance used has access to that accelerator locally
3. Use the online VSCode to develop and iterateThe workspace state is automatically snapshot as a docker image that is then used later on when I open the same project. This way I won't need to re-download all the required packages from scratch. I can also manually trigger the snapshot which can be used to rollback the workstation state to an older state in future.
I understand these are a lot of requirements but it would be a killer online IDE if all of this works
-
Hi Suvaditya - do you have a particular set of tools that you'd like us to consider providing "out of the box" support for?