Issues

• Community benchmark: Strata 0.1.41 vs llama.cpp's new MoE expert cache (--moe-cache-mib), same Coder IQ1_M files, V100-PCIe 32 GB + 16 GB RAM

  Niko1221/Strata#3082 · Filter by author christopherrobertbrooks-tech opened on Oct 18, 2026

• [Question]: GPU-access daemon for sandboxed assistants (bwrap/DSH) — wanted upstream?

  Niko1221/Strata#3081 · Filter by author ken2425pm opened on Oct 18, 2026

• serve: a tool call is parsed and returned even when the request declares no tools (content state, frontend.py:1285)

  Niko1221/Strata#3075 · Filter by author riverhh76 opened on Oct 18, 2026
  1

• [Windows/NVIDIA 0.1.45] engine 0xC0000005 at strata.exe+0x535c60 (Q4_K CPU dot product reads past a weight buffer), UD-Q4_K_XL with --batch-mtp on a RAM budget

  Niko1221/Strata#3071 · Filter by author dnnspaul opened on Oct 18, 2026
  1

• BENCH.bat report for Windows 11, RTX5080, 128GB RAM, i9-14900K

  Niko1221/Strata#3070 · Filter by author sroller opened on Oct 18, 2026

• Decode 2x slower under agent frontends (Windows, RTX 3060 12GB): GPU downclocks to P0 during MoE expert misses — locking clocks recovers ~60%

  Niko1221/Strata#3066 · Filter by author ahsvel opened on Oct 18, 2026

• [Feature]: Windows low-RAM mode (--mmap-experts): let STRATA_PROMPT_DIRECT work there - prompt reads are page-fault bound (opt-in patches and measurements)

  Niko1221/Strata#3065 · Filter by author As7656431 opened on Oct 18, 2026

• Community benchmark: R9700 vs RTX 5060 Ti — Qwen3.8-Flash-Next Q8_0 at 256K on HP Z4 G4; 3945WX server results

  Niko1221/Strata#3061 · Filter by author AloeKing opened on Oct 18, 2026

• Windows file tier: reading experts through the file cache after load gave +26% decode on RX 7800 XT (opt-in patch)

  Niko1221/Strata#3059 · Filter by author durikim1-cpu opened on Oct 18, 2026

• RTX 2080 Ti (sm_75) x2: STRATA_SM75_PERF=1 verified bitwise; prefill +46% from 0.1.43.1 to 0.1.45

  Niko1221/Strata#3056 · Filter by author jiuyinssf opened on Oct 18, 2026

• serve: lifetime token totals + a Codex-style daily activity heatmap in the monitor

  Niko1221/Strata#3053 · Filter by author wangzc1997 opened on Oct 18, 2026

• OCm 多进程共享注册主机内存崩溃问题 — 求助描述

  Niko1221/Strata#3051 · Filter by author weakspark opened on Oct 18, 2026
  1

• [Bug]: conversation-cache metrics skip events in log backlogs over 1 MiB

  Niko1221/Strata#3048 · Filter by author oscarlius opened on Oct 18, 2026

• [Bug]: completed non-finite recovery leaves stale conversation-cache metrics

  Niko1221/Strata#3047 · Filter by author oscarlius opened on Oct 18, 2026

• [Bug]: PleReader initialization failures can wait forever for nonexistent reads

  Niko1221/Strata#3045 · Filter by author j-luwierski opened on Oct 18, 2026

• [Bug]: strata-gguf --check returns success for invalid architecture and wrapped tensor bounds

  Niko1221/Strata#3043 · Filter by author j-luwierski opened on Oct 18, 2026

• [Bug]: refused batch admissions count as successful completions

  Niko1221/Strata#3040 · Filter by author oscarlius opened on Oct 18, 2026

• [Bug]: concurrent metrics readers duplicate conversation-cache events

  Niko1221/Strata#3039 · Filter by author oscarlius opened on Oct 18, 2026

• [Bug]: Windows/WDDM: parked-conversation snapshot copies corrupt at 262144 -> all-logits NaN (#879 symptom, separate trigger). One-file fix, verified on RX 9070 XT

  Niko1221/Strata#3036 · Filter by author denis-1337 opened on Oct 18, 2026

• [Bug]: Conversation-cache metrics remain stale after refused parking

  Niko1221/Strata#3033 · Filter by author oscarlius opened on Oct 18, 2026

• community test suite reports, Radeon 8060S (gfx1151) on a Ryzen AI Max+ 395 (Windows 11)

  Niko1221/Strata#3032 · Filter by author mistral13 opened on Oct 18, 2026

• serve: disk durability for parked conversations (saved conversations)

  Niko1221/Strata#3027 · Filter by author whitespades opened on Oct 18, 2026

• Tensor parallel (--tp 2) on 2× RTX 3080: patch attached, correct but not usable yet (~2 % of the layer split's prefill) — open for pickup

  Niko1221/Strata#3025 · Filter by author ZeaMQ opened on Oct 18, 2026
  1

• Official v0.1.45 runs NVFP4 but is ~17% slower than the strata-nvfp4 branch (type-40 experts not covered by EXPDEC/B6; no w4a4x2 prefill)

  Niko1221/Strata#3023 · Filter by author wangzc1997 opened on Oct 18, 2026

• [Bug]: gfx1100 2x RX 7900 XT layer split + --resident-experts (0.1.45,Q2_0):16K–100K prompt prefill stalls (watchdog #29), GPU reset / device lost

  Niko1221/Strata#3021 · Filter by author Wuqiyang312 opened on Oct 18, 2026
  5