Issues

• perf(server): keep CUDA graph replay when LoRA adapters are loaded

  superlinked/sie#840 · Filter by author svonava opened on Oct 20, 2026

• design(helm): define separation for operator-supplied dependency credentials

  superlinked/sie#780 · Filter by author krisztian-gajdar opened on Oct 17, 2026

• MUVERA: compute the encoding on the GPU, not on the inference thread's CPU

  superlinked/sie#443 · Filter by author fm1320 opened on Sep 30, 2026

• GLiClass: large separate-label-group requests can wait minutes in the queue under mixed load

  superlinked/sie#390 · Filter by author svonava opened on Sep 28, 2026
  3

• Batcher: a request that reaches an idle worker waits about 4 ms before dispatch

  superlinked/sie#373 · Filter by author svonava opened on Sep 25, 2026

• Model catalogue - larger open weight models and more powerful GPU

  superlinked/sie#230 · Filter by author ndjones opened on Aug 5, 2026
  3

• Your project is on StackMap — a curated map of the AI stack

  superlinked/sie#217 · Filter by author hoghweed opened on Jul 21, 2026
  2

• ask about this project

  superlinked/sie#191 · Filter by author TentenMarchhhh opened on Jun 12, 2026
  1

• Consider using pre-built Flash Attention kernels via kernels

  superlinked/sie#179 · Filter by author sayakpaul opened on May 28, 2026
  2