Issues

• prefill: the batched prompt path aborts (cublasGemmEx f16, cuBLAS status 14) on chunks below about 32 tokens

  Niko1221/Strata#1650 · Filter by author W1nge opened on Oct 9, 2026

• Fail to use strata on opencode due to long prompt.

  Niko1221/Strata#1648 · Filter by author RNGMARTIN opened on Oct 9, 2026

• setup.sh has no way of specifying gcc version

  Niko1221/Strata#1645 · Filter by author xobs opened on Oct 9, 2026

• Windows HIP (RX 7900 XT, gfx1100): every generation fails with "mtp prefill: unspecified launch failure" on a GSQ-RCO IQ3_XXS native pack (0.1.41 and 0.1.40.4)

  Niko1221/Strata#1644 · Filter by author johnmiu opened on Oct 9, 2026

• [Bug] v0.1.41 no-P2P peer prefill crashes on long prompts (RTX 2080 Ti + RTX 5050, CUDA Xid 31)

  Niko1221/Strata#1643 · Filter by author GarminChen opened on Oct 9, 2026

• args not support --reasoning-effort

  Niko1221/Strata#1641 · Filter by author ourfirebird opened on Oct 9, 2026

• Community benchmark: 5x Tesla P100 (sm_60), Flash-Next IQ3_S, 262K context — STRATA_STAGE_TRIM data + a Pascal Q6_K kernel

  Niko1221/Strata#1639 · Filter by author thedrzinger opened on Oct 9, 2026

• How to Install Two copy of Strata Systems Without Interference

  Niko1221/Strata#1634 · Filter by author K0KU20 opened on Oct 9, 2026
  4

• [Feature Request]: Support RAM-budgeted SSD-backed experts with multi-GPU layer splitting

  Niko1221/Strata#1631 · Filter by author KapellMeister61 opened on Oct 8, 2026

• Windows HIP (RX 7900 XTX): a long prompt with --prefill 8192 / auto makes the commit charge jump ~70 GB in one second, the engine dies at the commit limit (0xC0000409); 7500 and below stay flat

  Niko1221/Strata#1630 · Filter by author rekaXua opened on Oct 8, 2026

• [Linux, RTX 3090, stock 0.1.40.2] --ple-io direct on a DRAM-less NVMe reads the PLE table at 3.7-8.3 tok/s on the prompt path and trips the #29 watchdog; --ple-io mmap is 40-100x faster, first cold pass included

  Niko1221/Strata#1629 · Filter by author bluebirdlboro opened on Oct 8, 2026

• [Feature]: logit_bias on /v1/chat/completions (accepted today, but has no effect)

  Niko1221/Strata#1627 · Filter by author lukassbartik opened on Oct 8, 2026

• [Feature Request]: more of the llama outputs, etc

  Niko1221/Strata#1626 · Filter by author jeremyoha450 opened on Oct 8, 2026

• Support for gfx1152 is missing

  Niko1221/Strata#1625 · Filter by author nordlow opened on Oct 8, 2026

• Strata v0.1.41 Looping issues

  Niko1221/Strata#1622 · Filter by author XeonG opened on Oct 8, 2026

• serve: a cancelled request loses the conversation cache entry it restored (take() without a park), so the next request re-reads the prompt

  Niko1221/Strata#1620 · Filter by author taxah92 opened on Oct 8, 2026

• serve: a request waiting for the single request turn sends no bytes, so clients with a stream idle timeout abort it (and the retry re-reads the prompt)

  Niko1221/Strata#1619 · Filter by author taxah92 opened on Oct 8, 2026

• endless thinking loop

  Niko1221/Strata#1618 · Filter by author ilyatbn opened on Oct 8, 2026

• Auto layer assigns only 1 layer to card dropping prefill due to prompt-chunk clamp

  Niko1221/Strata#1616 · Filter by author tonydiep opened on Oct 8, 2026

• [Feature]: let harnesses recognize the context-overflow 400 (include a phrase they already match)

  Niko1221/Strata#1615 · Filter by author donnielrt opened on Oct 8, 2026

• Feature Request: Token repetition penalty

  Niko1221/Strata#1613 · Filter by author comet13579 opened on Oct 8, 2026
  1

• [Bug]: a tensor-less GGUF shard whose aligned data start is past EOF is refused (unsloth UD-Q5_K_XL shard 1)

  Niko1221/Strata#1611 · Filter by author CYoung83 opened on Oct 8, 2026

• Windows: the engine dies without a message (0xC0000409) when the commit limit is reached; the RAM checks only look at physical RAM

  Niko1221/Strata#1607 · Filter by author masel opened on Oct 8, 2026

• Intel Arc A770 (Alchemist, i915): prefill serialized (~47 tok/s, GPU idle, 1 CPU core), decode CPU-bound (all 14 cores at 90%, no VNNI)

  Niko1221/Strata#1606 · Filter by author Javalopes opened on Oct 8, 2026

• --batch-groups disables the per-slot conversation cache even when the group size is 1 (no pad row is ever written) → alternating long conversations re-prefill from token 0 every turn

  Niko1221/Strata#1604 · Filter by author 113636xfh opened on Oct 8, 2026