Hey HF!
A lot of times when I’m doing ML experiments with HF datasets and models, I’ll try something in a local notebook before wanting to scale it up.
That can be really painful. The DevOps of provisioning a GPU from a cloud, getting it setup with the right environment and moving my code over might take 5 minutes, and I have to do it every time I want to do a compute-intensive experiment.
That’s why I’m excited to share what I’ve been building:
It’s called Moonglow, and it lets you connect a local Jupyter notebook to a cloud GPU as easily as Google Colab, but without leaving your IDE or being locked into whatever GPUs Google offers. Instead, you can use the compute provider you prefer, whether it is a large cloud provider like AWS or a dedicated GPU reseller like Runpod.
You can try it out for free at moonglow.ai, and hopefully this makes your experimentation a bit easier! I’d also love any feedback if there are any issues with using Moonglow.