Issues

• extract: GLiNER2 entity and relation extraction silently read only the first 512 words

  superlinked/sie#450 · Filter by author svonava opened on Sep 30, 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

• Measure and improve single-stream decode throughput for Qwen3.8-27B-FP8

  superlinked/sie#435 · Filter by author svonava opened on Sep 29, 2026
  1

• Proposal: remote backends, serving a catalog model from an upstream

  enhancement
  superlinked/sie#415 · Filter by author krisztian-gajdar opened on Sep 29, 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

• GLiNER-family adapters: hardening follow-ups from #380 and #381

  superlinked/sie#382 · Filter by author svonava opened on Sep 26, 2026

• GLiFormer adapter: hardening follow-ups from #375

  superlinked/sie#378 · Filter by author svonava opened on Sep 26, 2026
  1

• Bundle images resolve ranged dependencies past uv.lock

  superlinked/sie#377 · Filter by author svonava opened on Sep 25, 2026

• GLiFormer: replay the static part of the forward as CUDA graphs

  superlinked/sie#374 · Filter by author svonava opened on Sep 25, 2026

• 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

• Release GLiClass CUDA graph memory under memory pressure before evicting models

  superlinked/sie#370 · Filter by author svonava opened on Sep 25, 2026

• gliner2 2.x changes gliner2-base-v1 and gliner2-large-v1 outputs (repeated NER mentions, structured extraction)

  superlinked/sie#369 · Filter by author svonava opened on Sep 25, 2026

• KEDA can scale a worker lane to zero while its model is actively loading

  superlinked/sie#293 · Filter by author igor-raits opened on Sep 17, 2026
  1

• 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