Nvidia Nemotron 3.5 Lightning and NeMo Switchyard

Re:Nvidia Nemotron 3.5 Lightning and NeMo Switchyard by halfdeadcat on Tuesday August 11, 2026 @08:17PM (#49266270)

Good luck getting it to run with NVFP4 on a DGX Spark, the very architecture Nvidia created that format for.

[ Parent | on HN ]

Re:Nvidia Nemotron 3.5 Lightning and NeMo Switchyard by Scaled on Wednesday August 12, 2026 @06:45AM (#49270353)

Can you elaborate on the problem? The article claims support for dgx spark

[ Parent | on HN ]

Re:Nvidia Nemotron 3.5 Lightning and NeMo Switchyard by rsolva on Wednesday August 12, 2026 @03:15PM (#49277260)

I have seen many a model card claim Spark compatibility, but without any proof or instructions. There are vibe-patched images of vLLM that kinda works with NVFP4 models, but I have not seen any support from NVIDIA. Does not seem to be a priority for some reason.

[ Parent | on HN ]

Re:Nvidia Nemotron 3.5 Lightning and NeMo Switchyard by Scaled on Wednesday August 12, 2026 @08:17PM (#49280296)

Thanks, I've been mulling over buying some of these so I appreciate the heads up.

[ Parent | on HN ]

Re:Nvidia Nemotron 3.5 Lightning and NeMo Switchyard by halfdeadcat on Friday August 14, 2026 @02:33PM (#49302819)

After thrashing on it for a day, I could not get it going on DGX Spark with FP4 quantization. I find this irksome, since Nvidia created the NVFP4 specifically for Blackwell. The Nvidia cookbooks for this model are all for H100. I tried vLLM, Ollama and various patches. As of right now on DGX: you can feasibly do FP4 on dense models. But FP4 + MoE is a largely a dead-end.

[ Parent | on HN ]