Issues

• Question on planning-stage optimization for VariableBlockSparseAttentionWrapper

  svg-project/Sparse-VideoGen#68 · Filter by author tie-pilot-qxw opened on Jan 16, 2026
  2

• Triton/CUDA Illegal Memory Access in triton_rmsnorm_forward with Wan2.1-T2V-1.3B-Diffusers

  svg-project/Sparse-VideoGen#55 · Filter by author 1717Li opened on Sep 29, 2025
  9

• Diffsynth Studio Support

  svg-project/Sparse-VideoGen#43 · Filter by author jxjx12138 opened on Jul 11, 2025

• A potentional issue on Blackwell GPUs (B200) for the placement operator

  svg-project/Sparse-VideoGen#40 · Filter by author wenhao728 opened on Jun 18, 2025
  1

• No obvious improvement observed on A100.

  svg-project/Sparse-VideoGen#38 · Filter by author Chuge0335 opened on May 27, 2025
  2

• No efficiency gains on A100

  svg-project/Sparse-VideoGen#36 · Filter by author jhl13 opened on May 23, 2025
  2

• The orig_wan implementation has a different block mask

  svg-project/Sparse-VideoGen#29 · Filter by author m-richa opened on May 6, 2025
  3

• Compatibility with other acceleration methods, such as TeaCache, Comfy-Wavespeed, torch.compile, and/or attention optimization

  svg-project/Sparse-VideoGen#12 · Filter by author Kazuki-Yoda opened on Apr 4, 2025
  5

• Does SVG Support Sequence Parallelism

  svg-project/Sparse-VideoGen#9 · Filter by author sywang0111 opened on Mar 25, 2025
  2

• Can this be converted into a Comfy node?

  svg-project/Sparse-VideoGen#3 · Filter by author CustodialMatter opened on Mar 13, 2025
  3