H100
H100-Qwen3-14B-MonEspaceSante-CPT-i1-GGUFPolyAgent_-_gemma-7b-ua-ckpt-h100-lower-lr-ggufH100-Qwen3-14B-MonEspaceSante-CPT-GGUFQwen3-8b-qwen-h100-20260909-CPT-merged-step-750PolyAgent_-_ua-mistral-0.3-ua-ckpt-h100-lower-lr-no-grad-accumulation-gguftinystories-fast-h100mixtral_i32000_o1024_bs4_tp1_H100_int8_weights_only-tllm_0.9.0.dev2024040200mixtral_i32000_o1024_bs6_H100_tp2-tllm_0.9.0.dev2024040200
Datasets
All datasets matching “H100”GKG-H100-Code-6Augfastvideo-wan-h100-sp1-nsys
FastVideo Wan2.1-T2V-1.3B — H100 SXM, SP=1 — Nsight Systems profiles (FA2 vs FA3)
NVIDIA Nsight Systems profiles of FastVideo
running Wan-AI/Wan2.1-T2V-1.3B-Diffusers text-to-video inference on a single
NVIDIA H100 SXM with sequence-parallel size 1.
The dataset contains two profiles of the same workload on the same machine,
differing only in the attention backend:
Attention backend
Kernel observed
File
FA2 baseline
FlashAttention-2
flash_fwd_kernel<...>… See the full description on the dataset page: https://huggingface.co/datasets/rich7421/fastvideo-wan-h100-sp1-nsys.tam-nfs-sys-u2204-rename-to-h100_rootpredictions-dev-staging-project-vajra-dev-staging-meta-llama-llama-3-8b-h100-a4e4fd59h100-multimodality-v0.1.0h100-offboard-209-20-156-204-20260711T141916Z