All pull requests

Pull requests

• gfx906/HIP build broken on v0.1.44.1: two compile blockers (platform/hip_compat/cublas_v2.h missing three cublas maps, and the new strata_mmq_w4a4 target missing GGML_USE_HIP)

  #2460·kietq8684-design opened Oct 14, 2026·

  ·

• engine: --dump-kv-select, the decode windows' KV block selections as a policy-bench trace (#2415)

  #2458·MrKop-git opened Oct 14, 2026·

  ·

• serve: an all-cached image turn never wakes the elastic encoder

  #2455·jerry78424 opened Oct 14, 2026·

  ·
  1

• setup: no KV streaming on one card that holds every expert and the whole KV cache (#2261)

  #2450·ischencheng opened Oct 14, 2026Contributor·

  ·
  1

• conversation cache: evict redundant holders without losing pinned prefixes

  #2443·KeyanHu-git opened Oct 14, 2026·

  ·
  1

• docs: community benchmark — RTX 5080 + 4070 Ti unequal pair, both orders (+ helper cache arm)

  #2439·leonininder opened Oct 14, 2026·

  ·

• SYCL: the IQ4_XS expert dot reads its codebook from registers

  #2437·deriknel opened Oct 14, 2026·

  ·

• SYCL: defer pipelined one-token commits to the host verdict (port of c81f9984)

  #2436·deriknel opened Oct 14, 2026·

  ·

• serve: an elastic image encoder that cannot start is a 503; its idle stop is logged on a layer split too

  #2435·i-Vladimir-i opened Oct 14, 2026Contributor·

  ·

• tools: the KV streaming page-pool policy bench - CLOCK against LRFU, LRU and Belady (#2415)

  #2434·MrKop-git opened Oct 14, 2026·

  ·

• SYCL: a pipelined window still on the GPU is polled again, not failed

  #2433·deriknel opened Oct 14, 2026·

  ·

• Add Intel GPU monitoring on Windows via IGCL

  #2429·marcelloc opened Oct 14, 2026·

  ·

• bench: RTX 5060 Ti 16 GB on Ubuntu 24.04, 0.1.44.1, Flash-Next IQ3_XXS

  #2428·liulib opened Oct 14, 2026·

  ·

• HC_Q8: skip the layers' bf16 hyper-connection copies on a layer split too

  #2427·paltian opened Oct 14, 2026·

  ·

• Layer split auto: STRATA_SPLIT_UNIFORM=1 restores the uniform miss cost

  #2425·MEsfits opened Oct 14, 2026·

  ·
  1

• serve: GET /metrics gives each request's PCIe expert bytes and its own KV streaming share, and a bandwidth summary

  #2423·MrKop-git opened Oct 14, 2026·

  ·

• web: the PCIe metric shows the current link generation, not the card's capability

  #2422·MrKop-git opened Oct 14, 2026·

  ·

• Video: the encoder reads the clip itself (ENCV - mtmd samples its frames at video_fps and merges them)

  #2421·miaomiaozii opened Oct 14, 2026·

  ·

• arm64: build + run Strata on NVIDIA DGX Spark (GB10, sm_121) - rebased onto v0.1.41, IQ3_S validated

  #2419·ccatzin-zumwalt opened Oct 14, 2026·

  ·

• bench: community report, 2x RTX 5060 Ti 16 GB, Xeon E5-2690 v4, IQ3_XXS at 204,800-token context, engine 0.1.44.1

  #2414·R6DJO opened Oct 14, 2026Contributor·

  ·

• docs/AMD_HIP.md: short GPU stalls while the kernel compacts memory, and how to keep them out of a speed comparison

  #2413·libratechw opened Oct 14, 2026Contributor·

  ·

• fix: avoid CUDA OOM during conversation-cache restore

  #2409·MarsLuay opened Oct 14, 2026·

  ·
  1

  1

• bench: community report, RX 9070 XT 16 GB on Windows, IQ3_XXS, engine 0.1.44.1

  #2405·winky1717 opened Oct 14, 2026·

  ·

• Community benchmark: RTX 4080 SUPER 16 GB, Ryzen 7 7800X3D, 64 GB, IQ3_S at 192K (0.1.44.1)

  #2404·olituks opened Oct 14, 2026·

  ·
  1

• Add opt-in usage accounting, output continuation and authenticated remote API

  #2400·QuickkApps opened Oct 14, 2026·

  ·
  1