I use Jupyter a lot for ML research, and something I find really annoying is that switching between my local computer to test stuff quickly and a remote GPU to scale up my experiments can take 5 minutes: provisioning a GPU, SSHing into it,
setting up the environment, pulling my notebook over etc.
Colab is a decent option, but I like my IDE and I often need GPUs that Google doesn’t offer. That’s why I built Moonglow:
It lives in VS Code and lets you connect your local Jupyter notebook to cloud GPUs (using your API keys) just by changing the kernel runtime!
It’s free to try out at moonglow.ai, and I’d love to know if this is helpful or if there are any issues you might have when using it!