Temas / Documentation
Documentation
Auto topic: Documentation
Issues
- #5949 [Engine] HIP/Linux: make HSA_USERPTR_FOR_PAGED_MEM=0 the default on discrete AMD cards
- #6531 [Benchmark] gfx1100 (RX 7900 XTX, IQ3_S): existing switches give 1.9–3.7x prompt and 1.2x decode speedup
- #6438 [Engine] Strata 2.17 + Lazarus 4.8 + Lazarus MCP Assistent not working ?
- #6525 [Engine] Drop in MTP head replacement trained against quantized model output: +3.7 / +4.4 % decode at --spec 3
- #6511 [Benchmark] 2x Tesla P40 (sm_61) split: 0.2.17 gives +9% decode, but 220K prefill is still ~5% under the 0.2.13.1 MMQ path (same-day A/B)
- #6273 [Engine] 16 GB cards: guard the ~1 GB VRAM floor during KV growth and document the per-format context limit
- #6232 [Engine] 384K compaction re-reads 264K tokens (456 s): content-addressed, position-independent segment cache for near-zero-reread compaction
- #6327 [Engine] Compaction re-reads 331K-342K tokens five times (19.5-20.1 min each): the cache budget is the blocker (PR #4338)
- #6496 [Other] Windows HIP rdna2 package (0.1.63) works on a real RX 6900 XT (gfx1030)
- #6429 [Benchmark] hipBLASLt tuning table by default on gfx1100 (RX 7900 XTX, IQ3_S): speed, decode and logits
- #6462 [Benchmark] RTX 4090 (sm_89): sm_86 bit-exact set +6%, sm_120 switches +4% on UD-Q4_K_XL, CPU share 32/6144 cuts 3K-token follow-up turns 26%
- #6448 [Engine] Scope: plan an integrated Intel GPU (Arc B390, Panther Lake) as one pool of memory with the system in the SYCL engine
- #6121 [Engine] Non-finite logits (#879) with Unsloth UD-Q4_K_XL native on RTX 4090 — deterministic at 267 tokens; IQ4_XS and Q5_K_XL clean on same engine
- #5995 [Engine] Exit code 0xC0000409: stack buffer overrun on V100 + RTX A4500
- #6413 [Benchmark] v0.2.17 on 2x Arc Pro B60 (layer split) and 1x Arc Pro B70 / Qwen3.8-Flash-Next UD-Q4_K_XL / 262K: same speed as v0.2.16, `--pipeline-windows 2` no longer produces `!!!!` (12 of 12)
- #5551 [Engine] q36 on the qwen35moe 19B-A3B shape: three localized walls with clean data
- #5876 [Engine] KV reservation before expert placement starves the expert cache on small cards
- #6400 [Engine] + [Docs] SYCL setup & LXC/multi-GPU flow is a complete mess (paths, build dir, device selection)
- #6156 [Engine] Qwen3.6-35B-A3B on 2x RTX 3060: qwen35moe ignores multi-GPU flags — patch adds a card-1 expert tier, 37.6 → 74.8 tok/s (+97%)
- #5781 [Engine] Support N+ dedicated MoE expert-cache helper GPUs
- #5756 [Engine] Tool-call value truncated when the model skips a parameter's own </parameter> after a nested pair (Qwen3.8-Flash-Next)
- #6269 [Engine] gfx103x: prompt throughput gaps left by the MMQ removal (dense projections, IQ4_XS down)
- #6203 [Engine] serve/server.py engine_args() ignores hand edits to args without user_args
- #6174 [Docs] Mac: source setup gives Q2_0 the DFlash2 drafter, but MAC.md says only the two Unsloth sets get it
- #6086 [Other] Terminal: shell tab completion for the strata command
- #5996 [GUI] helpers on one card: three read-only helpers ran 163, 266 and 351 minutes, with no cost warning and no ceiling
- #5991 [Other] the shipped Harness 0.2.1-alpha.3 has no model-switch pressure check, added upstream in dsh-v0.2.1-alpha.4
- #5807 [Engine] DFlash2 drafting confidence is mis-tuned for a 16 GB Ada card: `CONF_STEPS` chooses 0.8, the measured optimum is 0.5, worth +16.7 % decode
- #5778 [GUI] Scope opt-in side questions, deeper research and private workflow skills
- #5612 [Engine] conversation cache: on a layer split, a sub-agent that matches only a shared head takes the whole parked conversation and drops its checkpoints
- #6236 [Other] RTX 4090 / Qwen3.8: 131K Context, Performance & CUDA Stability — Field Report
- #5557 [Engine] SYCL: weight arena is ONE USM device allocation — returns NULL silently on devices with a max-alloc cap (WSL2 GPU-PV, 1 GiB measured); crash 'cudaMemcpy failed for output_hc_down.weight'
- #6235 [Other] with 2.16 update.sh is missing.
- #6364 [Engine] Prefer FP32 TLA over INT8 ESIMD for Battlemage prompt attention
- #6361 [Benchmark] Unequal NVIDIA pair, both card orders, `--layer-split auto`, and the helper mode: RTX 3060 12 GB + RTX 5060 Ti 16 GB, 0.2.16, NVFP4-GPTQ
- #6355 [Engine] Layer-split serves cannot reach either disk cache: the SSD spill tier and session files are single-serve only
- #6354 [Benchmark] RTX 3060 12 GB on an AVX-only Xeon E5-2630 v2 (no AVX2/FMA): 0.2.16 (engine 0.1.62) against the 0.2.0 report, IQ2_XS and NVFP4
- #5549 [Benchmark] Single vs dual rx9070 (hip): engine 0.1.58, ud-iq4_xs-v31
- #6248 [Engine] Retracted (reporter config error); kept: per-request strata_tune no-op on 0.1.62, boot-time pcie_frac costs 2-3x on SM86
- #6322 [Engine] v0.2.16 source build on gfx1201 segfaults before any output with system ROCm 7.16; STRATA_HIP_QUEUE_PRIME=0 and the official rx9070 package are unaffected
- #6279 [Engine] RTX 4090 (Windows, CUDA 13): decode about 10-20% slower on 0.1.62 vs 0.1.60 with UD-IQ4_XS-v31 (prompt reading unchanged)
- #6317 [Engine] Scope for an opt-in Vulkan research path on the Intel Arc B390 iGPU
- #6306 [TUI] Unable add a mcp server to "strata launch" on Linux
- #6295 [Benchmark] v0.2.16 on 2x Arc Pro B60 (layer split) and 1x Arc Pro B70 / Qwen3.8-Flash-Next UD-Q4_K_XL / 262K: same speed as v0.2.13, `!!!!` with --pipeline-windows 2 back to 0 of 12
- #6296 [Other] 0.2.16 layout: Intel (SYCL) setup cannot start the server ("cannot open .../engine/run-*.sh"), and `strata bench run` from a git clone is always nonconforming
- #5927 [Engine] a pack without tokenizer.json is silently encoded with the qwen35 pre-tokenizer (DeepSeek digits split into characters)
- #5908 [Engine] DeepSeek-V4.1 Flash SYCL on B70 — environment and performance
- #5663 [Engine] ds4 won't start on R9700
- #5624 [Engine] DeepSeek-V4.1-Flash 2.25 bpw on one RTX 4090: two settings worth productizing, the DS4_* knobs, and one start-up peak we cannot explain
- #5570 [Engine] DeepSeek-V4.1 has no --no-prefill-borrow: on a 16 GB card the prompt loan costs half the expert cache
- #6268 [Engine] The no-VRAM-left-for-expert-cache error never suggests --kv-grow, only quality-reducing fallbacks
- #5858 [TUI] Add a user-wide configuration file shared across models and presets
- #5968 [GUI] Bypass mode has no way to obtain Windows administrator rights: the sandbox ladder tops out at the app's own token
- #6234 [Other] RTX 4090 + Qwen: 131K Context, Performance, CUDA Stability — Community Field Report
- #5674 [Engine] Metal: --batch 2 --batch-mtp exits on the first concurrent pair ("two rows of one batch window are the same slot")
- #5818 [Engine] Mac (M2 Ultra): Unsloth UD-IQ1_M (v3.1) generates garbage text; llama.cpp on the same files is coherent
- #6013 [Other] --ple-gguf + Unsloth engram table: the manual swap that works today (re-stamp the header to strata-ple; verified UD-Q4_K_XL, engine 0.1.59.1)
- #6090 [Benchmark] RX 9070 XT (gfx1201) + DFlash2: two weeks >60 token/sec on Strata 0.2.10 / 0.1.56
- #6003 [Benchmark] Post-merge field report for #126 (Task Scheduler expert load): defaults vs Priority 4 + HighestAvailable vs interactive on a Windows node
- #5928 [Benchmark] CMP 170HX (sm_80, modified 64 GB) / Qwen3.8-Flash-Next UD-Q4_K_XL / 256K / 0.2.13.1
- #5857 [Benchmark] 2× RTX 3060 12 GB, vfio-passthrough KVM VM — qwen3.8-flash-next iq3_s, layer split
- #5762 [Benchmark] DFlash2 on gfx1201 (R9700, 32 GB): no prompt stall on 0.1.59.1; window 8 is what pays, conf 0.35 costs 10%
- #5759 [Benchmark] Arc Pro B70 (x16) + B60 (x8) layer split on EPYC 7443: IQ3_S/IQ2_XS at 262K, engines 0.1.54-0.1.59.1
- #5683 [Benchmark] the machine line names a disabled iGPU and 2.0 GiB of VRAM instead of the two cards the engine used
- #5680 [Benchmark] 2026-11-09-community-rtx-5070-plus-rtx-5060-ti-qwen3-8-flash-next-iq3-s
- #5666 [Benchmark] RX 7900 XTX (gfx1100) / Strata 0.2.13: the Linux Docker image without the gfx1100 hipBLASLt table reads 16K prompts 35 % slower; with the table equal to 0.2.12
- #5610 [Benchmark] Independent distributional quality: Strata vs llama.cpp vs BF16 truth, 399,600 teacher-forced positions
- #5607 [Benchmark] Radeon 780M (gfx1103): DeepSeek-V4.1-Flash IQ1_M on strata-dsv41, HIPK +80 % prompt, --dspark on +70 % decode
- #5583 [Benchmark] 2x Arc Pro B60 (layer split) and 1x Arc Pro B70 / Qwen3.8-Flash-Next UD-Q4_K_XL / 262K / v0.2.12
- #6204 [Engine] opt-in merging of consecutive text/thinking deltas on /v1/messages
- #6083 [Engine] Single-socket dual-CCD Zen 5: the 12 pool workers pin to CPUs 0-12, the second CCD stays idle, and the loaded die hits 95 °C where an all-core load runs 72 °C
- #6094 [Engine] gfx103x layer split on PCIe 3.0: STRATA_PREFILL_CPU_SHARE=auto with MAX=16384 reads prompts 1.26-1.54x faster (2x RX 6800 XT)
- #5796 [Engine] Tenstorrent Blackhole p150a card(s) as an engine
- #5899 [Engine] Offer: XeMaya Intel SYCL model paths and optimizations for upstream integration
- #5867 [Engine] Pipelined batch slots: a slot that reads alone sends its next window ahead through a group that stands idle
Pull requests
- #5699 [Engine] session-identity: the whole STRATA_* environment and the device plan join the config fingerprint (#5379)
- #4310 [Engine] session files: format v2 with per-stage save/restore under --layer-split
- #6520 [Engine] Inkling-2-Mini from the official NVFP4/MXFP8 checkpoint, expert offload and CPU experts for 8-16 GB cards
- #6023 [Engine] peer_stream_ring: fix >= 2K peer_i8 ring wraparound + V100 docs
- #6210 [Engine] Opt-in P2P q8_1 decode push, fused P2P down-projection, and GPU peer sync flag
- #6529 [Engine] Volta sm_70 decode optimizations, zero-VRAM in-place Q8_0, NVLink split head, and volta-forever detection
- #6451 [TUI] strata launch --model-path: name Swift's shard 1 as --ple-gguf, where its n-gram table is
- #4307 [Other] serve+webui: unload that works — pre-staged checkpoints, polite waits, honest answers
- #4306 [Other] serve: session ladder — conversations survive an engine unload
- #3552 [Engine] skip file-tier experts while a helper GPU is attached
- #6535 [Benchmark] RTX 5060 Ti / Qwen3.8-Flash-Next / IQ3_S / strata-qwen38-thinking-v1-draft.1
- #6532 [Engine] prefill: opt-in STRATA_I8_DEFER - unpicked streamed entries no longer cut the int8 expert groups into fragments (gfx1030 4K prompt +3.6%)
- #5897 [Engine] serve: let the "vision" section carry its own "env" for the image encoder
- #6506 [Engine] add Inkling 2 model family and dual-GPU looped MoE engine support
- #3607 [Engine] native Windows support for Intel Arc SYCL, rebased onto 0.2.18 (engine 0.1.64)
- #5820 [Engine] Multiple API keys with per-key concurrency limits and reserved capacity
- #6365 [Engine] Prefer FP32 TLA sparse prefill on Battlemage in builds with local TLA headers
- #6514 [Benchmark] RX 7600 XT (gfx1102): ROCm 7.2.4 vs 10.1, and the CachyOS ROCm 10.1 crash
- #6510 [Engine] --experts-wex4: read the routed experts from a checked WEX4 container (opt-in, file tier)
- #6468 [Benchmark] RX 9070 XT on Windows, Gyro-S at 131K: app 0.2.17 / engine 0.1.63 with a tuned config (standard 12/12)
- #6176 [Engine] q4u32: the 4-bit KV cache with a fitted step as the default on RDNA4
- #6498 [Benchmark] RTX 3060 12GB / Qwen3.8 Flash-Next Q2_0 / standard v1 draft.1 + 260K xhigh
- #6486 [Engine] serve: an mcp__ prefix added to a plain tool, or dropped from an MCP tool, is that tool
- #2965 setup: a separate opt-in layers 8-33 projection vector for the original Qwen3.8-Flash-Next (any quantization)
- #6477 [Benchmark] Arc Pro B60 / Qwen3.8-Flash-Next / IQ3_S / strata-qwen38-thinking-v1-draft.1
- #3957 [Engine] serve: a long conversation survives an engine restart - parked conversations saved at a graceful stop and loaded at the next start, to files or (Windows) kept in RAM with no disk writes (opt-in STRATA_PARK_DIR, STRATA_PARK_RAMLIKE=1; layer split included)
- #5476 [Engine] gfx103x decode: opt-in one-warp MMVQ for K <= 4096 and a per-format expert layout table (STRATA_MMVQ_NW1, STRATA_EXP_TABLE)
- #5475 [Engine] opt-in int8 copies for the persistent hc read (STRATA_HC_PERSIST_Q8, STRATA_HC_I8ONLY)
- #6366 [Engine] Build the Battlemage TLA prompt attention by default from a sycl-tla submodule
- #2843 [Engine] Slot checkpoints for clients that rewrite the end of their history (3/3, depends on #2527 and #2528): STRATA_SLOT_CHECKPOINTS, STRATA_BRANCH_CHECKPOINT, BCHECKS
- #2528 [Engine] Slot placement policy for STRATA_SLOT_CACHE_AFFINITY (2/3, depends on #2527): protection by history length, retention scores, value decay, no wait
- #2527 [Engine] Keep conversations in pipelined batch slots (1/3): active-row windows, park=, cache-aware placement behind a policy interface
- #1614 [Engine] shared KV pool for all lanes (--kv-pool-tokens): at 2 x 256k the first token comes 2-5x sooner and lane moves ~10x faster; off = identical to main
- #6465 [Engine] bound host RAM during conversation swaps
- #6464 [Benchmark] Strix Halo 64 GB unified memory (Ryzen AI Max+ 395)
- #6453 [Engine] SYCL: an integrated Intel GPU is planned as one pool of memory with the system (Arc B390)
- #6452 [Engine] serve: an on-demand CPU picture reader starts without asking the expert cache for VRAM
- #6383 [Docs] add reusable project workflow skills
- #6458 [Engine] STRATA_FILL_STAGED: copy mapped blobs to the GPU via pinned buffers
- #6455 [Engine] SYCL: an Xe3 GPU (Panther Lake, Arc B390) runs the Battlemage kernels; STRATA_INTEL_GEN sets the class
- #6454 [Engine] serve: an integrated Intel GPU's memory is the system RAM in the Monitor and the free-memory checks
- #4334 [Engine] STRATA_HOST_TIMELINE host-timeline split + summary-time draft-profiler flush (opt-in, measurement only)
- #3919 [Engine] kv: size the resident ring from the run's depth (opt-in, STRATA_KV_RING_ADAPT)
- #3648 [Engine] --peer-device host route: STRATA_PF_PEER_PIECES=1 (opt-in) moves the activations and the sums in 8 pieces (+7 % prompt on 2x V100)
- #2427 [Engine] HC_Q8: skip the layers' bf16 hyper-connection copies on a layer split too
- #4766 [Engine] gfx906: build dflash2 for Radeon VII
- #6430 [Engine] q36: --kv fp32|fp16|fp8|fp4 for the K/V cache — 262,144 tokens on one 12 GiB card
- #6433 [Engine] Fix: read the MTP draft layer's expert count from experts.bin, to support models with other expert counts such as pruned models
- #2784 [Engine] HIP: the RX 6600 / 6650 XT (gfx1032) in setup's cards
- #6263 [Engine] Add opt-in MPI/NCCL split inference for two DGX Sparks
- #6416 [Engine] serve: root the full-suite CI step at terminal/ so serve tests import
- #4374 [Other] serve: opt-in "system_reminders": "tail" keeps a client's system-reminder blocks at the end of the prompt
- #6414 [Engine] gfx906: reference build for MI50/MI60/Radeon VII (0.2.11 + int6 kernels)
- #6314 [Benchmark] RX 6700 XT and RX 6600 (x1, x2, with and without a rocBLAS table) / Qwen3.8-Flash-Next / IQ3_S / strata-qwen38-thinking-v1-draft.1 on the 0.2.16 rx6000 package
- #4859 [Engine] a 3-bit Lloyd-Max K/V (--kv oct3) with an exact fp16 window, and the gfx1100 decode ports
- #6408 [Docs] Community benchmark: one Radeon VII — ROCm 10.1 vs 7.14, engine 0.1.63 vs 0.1.62
- #5494 [Other] Windows: recognise the Radeon Pro V620 (PCI id 0x73A1) as gfx1030
- #4272 [Engine] Metal: opt-in routed-expert tiles (STRATA_METAL_MOE_GU/_DN); Q2_0 8,1,2 + 4,2,4 is 17-24% faster on M1 Max
- #6166 [Engine] opt-in blob verifier for the direct expert path: STRATA_BLOB_VERIFY
- #6260 [Docs] Explain DFlash2 selection during Metal setup
- #6249 [Engine] Opt-in STRATA_PREFILL_EQUAL_TAIL: keep a short last prompt chunk instead of a second streamed pass
- #6228 [Engine] split: Gyro trellis experts (TQ2_T, TQK6, TQK7) on Turing (sm_75), opt-in
- #6087 [Other] Terminal: strata complete prints bash/zsh/fish tab completion from the CLI's own grammar
- #6009 [Docs] speed projection: a request that switches reads its prompt again, the other setting's parked conversation stays (measured on 2x P100)
- #5835 [Engine] Conversation cache: an opt-in disk tier for evicted, idle and oversized conversations (--conversation-cache-dir)
- #5599 [Docs] reproducible output: add the CPU prompt share and the route tail skip, and correct both rows (#4673)
- #5025 [Engine] hip: gfx1151 hipBLASLt rows for Qwen3.6 prompt projections
- #4651 [GUI] add --resume (STRATA_RESUME=1) to the Linux AppImage so your last model loads as Strata starts
- #4582 [GUI] add --lan (STRATA_LAN=1) to the Linux AppImage so the model is on your network as Strata starts
- #4295 [Engine] gfx1151 hipBLASLt table for a system ROCm 7.2.4 (hipBLASLt 100202)
- #4232 [GUI] API monitor: show the tokens held in the engine's KV cache
- #4154 [Docs] ROCm 6.3.3 / gfx906 report — 2x MI50, system ROCm, build + run + 12 h split-config soak
- #4153 [Other] setup: recognize the Instinct MI50 / MI60 / Radeon VII (gfx906) as a supported card
- #3348 [GUI] Add an iPhone companion with shared chats and model downloads
- #2645 [Engine] Metal: opt-in fp16-slab expert path (mpp) for streaming prompt chunks