Issues

• Feature Proposal: Accelerating NVMe Weight-Streaming via Block-Parallel Speculative Decoding (DFlash2 Style)

  sqliteai/warp#78 · Filter by author sslupsky opened on Oct 5, 2026

• DeepSeek-V4.1 leaves 15% on the table: the i8mm trunk kernel is built, measured and off by default

  enhancement
  sqliteai/warp#70 · Filter by author marcobambini opened on Sep 17, 2026

• A Qwen load sets the trunk kernel for the whole process, so a Kimi context already open switches to i8mm

  bug
  sqliteai/warp#68 · Filter by author marcobambini opened on Sep 16, 2026

• the drive link for a converted glm53 waste

  sqliteai/warp#67 · Filter by author neilbauman21-hub opened on Sep 15, 2026

• GLM 5.3

  sqliteai/warp#58 · Filter by author johnbarbaro737 opened on Aug 30, 2026

• A reservation ledger, for the case where the floors fit and the caches do not

  sqliteai/warp#49 · Filter by author marcobambini opened on Aug 24, 2026
  2

• Gate 7: Strix Halo (128 GB, Zen 5 AVX-512/VBMI) dataset for the fine cache sweep

  help wanted
  sqliteai/warp#37 · Filter by author Lrrr908 opened on Aug 13, 2026
  7

• Ship weights on IPFS or the web

  sqliteai/warp#24 · Filter by author gkgoat1 opened on Aug 4, 2026
  4

• CUDA backend: what would have to be true for it to pay

  enhancementhelp wanted
  sqliteai/warp#11 · Filter by author marcobambini opened on Aug 1, 2026
  17