Issues

• 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

• 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

• AMD GPU 9060 XT is not suppported

  Niko1221/Strata#3019 · Filter by author lumylovepandas opened on Oct 18, 2026
  4

• qwen35 (Qwen3.6): --mtp N is silently inert in serve mode unless --batch N --batch-mtp are also passed; setup's own config omits them

  Niko1221/Strata#3009 · Filter by author FlyoverBlind opened on Oct 18, 2026

• qwen35 (Qwen3.6): --mtp N is silently inert in serve mode unless --batch N --batch-mtp are also passed; setup's own config omits them

  Niko1221/Strata#3008 · Filter by author FlyoverBlind opened on Oct 18, 2026

• 这个性能指标咋样

  Niko1221/Strata#3007 · Filter by author phuai007 opened on Oct 18, 2026
  1

• [Bug]: The Mac (Metal) engine ignores a request's sampling: temperature, top_p, top_k, min_p, seed and the penalties do nothing, logit_bias is dropped, and every answer is the greedy one

  Niko1221/Strata#3006 · Filter by author oceanio opened on Oct 18, 2026

• getting errors on 0.1.45 : registration error

  Niko1221/Strata#3005 · Filter by author Schumi08 opened on Oct 18, 2026

• bench: community report, 4x Tesla T4 (sm_75, Linux, Docker): UD-Q6_K_XL on 0.1.45, +17% decode vs 0.1.42, 128K context verified with a 93K prompt, and the per-request VRAM warning that --expert-cache auto sets off by itself

  Niko1221/Strata#2999 · Filter by author ericoh88 opened on Oct 18, 2026

• KV cache q8_0 + q5_1

  Niko1221/Strata#2992 · Filter by author adbensi opened on Oct 18, 2026