This is NVIDIA Nsight Systems profiling traces for SC26 Paper "Demystifying Scale-Up Fabric Choices in Multi-GPU Systems for LLM Inference Serving". Please use NVIDIA Nsight Systems 2025.5.1 or newer
This is the Llama2-70B Model in SafeTensor format. Originally published in the HuggingFace repository, the mirror on Zenodo is used to preserve the artifact accompanying the SC26 Paper "Demystifying S