I am very new to this but have a DGX spark that I am installing Nemoclaw on. During the installation you get the option to install NIMS (which I understand it local Inference) vs NVIDIA cloud. I try to install NIMS with the Nemotron 70B LLM and it gets all the way to the end and then doesn’t install and fails over to the NVIDIA Cloud option. Is anyone else having this issue? Is it known? Is NVIDIA working on it? Sorry if this is the wrong forum. I didn’t see one for the Nemoclaw by itself.
At the end of the day, can someone articulate the really benefit of running this versus vanilla OpenClaw? Particular outside of an enterprise environment that requires lots of controls?
After some investigation, I can tell you that NIMs are not currently supported in NemoClaw. For local inference, Ollama is what is available at the moment. If you need to use a NIM, you would need to use OpenShell directly to setup an inference provider and launch OpenClaw sandbox with OpenShell.
However, NIM support in NemoClaw is coming in the future.