Unlocks a big speed up in ML workloads! Up next, WASM :)
Only nvidia?
Nope! Basically anything that you can mount into a container, we support. We just pass it through. You have to install the drivers on the node first though.
Disclosure: I co-founded Bacalhau
Argh my mistake! I thiught we were able to get it in, but it turns out we fell back to just using the NVidia docker work. You can achieve this in other ways (using the experimental pluggable executors), but not in a container.