Issues

• 0.1.41: 128K decode is 18% slower than 0.1.40.3 on an RTX 4090 (sm_89); STRATA_MMVQ_IL=0 gives all of it back

  Niko1221/Strata#1842 · Filter by author Wanderingss opened on Oct 10, 2026

• [Feature]: resident RAM mode — let the RAM copy follow the conversation when it cannot hold every expert (opt-in, PR #1838)

  Niko1221/Strata#1837 · Filter by author SanghunYun opened on Oct 10, 2026

• How/when are the security issues discussed here fixed?

  Niko1221/Strata#1835 · Filter by author Coderx7 opened on Oct 10, 2026
  1

• Windows HIP: START-HERE.bat --calibrate works on RX 7900 XTX (gfx1100) - ~15% decode gain

  Niko1221/Strata#1831 · Filter by author twining69-tech opened on Oct 10, 2026

• Chinese output collapses into a token loop after a few hundred tokens

  Niko1221/Strata#1827 · Filter by author beginor opened on Oct 10, 2026
  1

• Please include "repeat_penalty" in strata-xxx.json "sampling" section

  Niko1221/Strata#1819 · Filter by author avkwok opened on Oct 10, 2026

• the reply repeated one token ('!') 256 times in a row:

  Niko1221/Strata#1815 · Filter by author rawdraw2026-tech opened on Oct 10, 2026

• Model writing a literal </think> emits the special token: quoted tag ends the reasoning mid-sentence and the turn stops (finish_reason=stop)

  Niko1221/Strata#1814 · Filter by author danirebollo opened on Oct 10, 2026
  2

• [Feature]: a queue cap that answers 429 + Retry-After instead of queueing without end

  Niko1221/Strata#1809 · Filter by author streber42 opened on Oct 10, 2026

• setup.py get_llama_cpp(): a truncated llama.cpp zip is marked done, then fails with BadZipFile on every rerun

  Niko1221/Strata#1797 · Filter by author agnaylor opened on Oct 10, 2026

• Docs: on xe there is no SSD expert tier, and --spec 0 is refused for native packs, but the Intel docs and TROUBLESHOOTING say otherwise

  Niko1221/Strata#1796 · Filter by author agnaylor opened on Oct 10, 2026

• sycl/CMakeLists.txt test registration lags the top-level one: iq_parity and ple_parity fail on every checkout

  Niko1221/Strata#1795 · Filter by author agnaylor opened on Oct 10, 2026
  1

• setup_intel.py: sycl_version() returns "0", so setup refuses UD-IQ4_XS and UD-Q4_K_XL on Intel

  Niko1221/Strata#1794 · Filter by author agnaylor opened on Oct 10, 2026

• bench: RTX 5090 IQ3_S 262K, honest baseline 160-195 tok/s

  Niko1221/Strata#1791 · Filter by author Dylan997S opened on Oct 10, 2026

• [Bug] Windows HIP (Radeon AI PRO R9700, gfx1201): engine hangs at verify-window capture at --max-context 262144 — LiveKernelEvent 141 (TDR); 229376 and below start and serve fine

  Niko1221/Strata#1785 · Filter by author Metalazzo opened on Oct 10, 2026

• SYCL: "effort_position": "end" never turns on - the server looks for --tail-role-token in the launcher script, not the engine binary

  Niko1221/Strata#1781 · Filter by author leadgtr7 opened on Oct 10, 2026

• Web chat interface cannot render mathematical expressions

  Niko1221/Strata#1780 · Filter by author yutils opened on Oct 10, 2026
  1

• Community benchmark: Tesla V100-32GB, i5-10505 (6-core Comet Lake), 32 GB RAM

  Niko1221/Strata#1776 · Filter by author xWaLeEdOoOx opened on Oct 10, 2026
  1

• native PLE postops launch: no kernel image is available

  Niko1221/Strata#1772 · Filter by author savergiggio opened on Oct 9, 2026

• Latex support

  Niko1221/Strata#1769 · Filter by author LeFaucheur0769 opened on Oct 9, 2026
  1

• SYCL --batch 2: a prompt read in parts while another slot decodes leaves non-finite residuals (prompt-path cache-slot loan), unless --no-prefill-borrow

  Niko1221/Strata#1762 · Filter by author leadgtr7 opened on Oct 9, 2026
  2

• Regression: automatic GPU ordering in be8db787 reduces decode throughput ~15% on dual RTX 5070 Ti

  Niko1221/Strata#1760 · Filter by author L-proger opened on Oct 9, 2026
  1

• [Feature]: fail the test build when a release-engine CUDA kernel spills registers or uses local memory (per sm_)

  Niko1221/Strata#1758 · Filter by author Avicennasis opened on Oct 9, 2026

• Layer split (RTX 3090 + 2x Tesla P40, Windows): every batched prefill chunk aborts with prefill gemm: cublasGemmEx: cuBLAS status 14 — 100% on 0.1.40.4 / 0.1.41, single-GPU unaffected

  Niko1221/Strata#1757 · Filter by author vgame86 opened on Oct 9, 2026

• [Bug]: CPU prefill share: a failed cudaHostAlloc kills the engine instead of falling back to the GPU-only path (single GPU, Windows, 64 GB host)

  Niko1221/Strata#1756 · Filter by author noderex opened on Oct 9, 2026