Qwen3.8-Flash-Next on any consumer hardware: one-click install for Windows / Linux. Strata inference engine, OpenAI/Anthropic API on localhost, optional image input.
C++
docs: the xe expert tier has no SSD fallback under the wrapper default, and a native pack needs --spec 2 or more (#2153, #1796)
Yuu05c1dabf1aab
diff --git a/docs/INTEL.md b/docs/INTEL.mdindex 156150390..85b501da3 100644--- a/docs/INTEL.md+++ b/docs/INTEL.md@@ -127,9 +127,13 @@ the oneAPI runtime image with the serve pipes attached. Paths in the config's `a"sampling": {"temperature": 0.6, "top_p": 0.95, "top_k": 20, "repetition_penalty": 1.05}}```-With `--stream-experts` any expert left out of VRAM is read from the SSD and computed on the CPU when routed, so the reserve-matters: on the B70 with Coder IQ1_M a 1,536 MiB reserve left the cache 128 experts short and decode fell to 5-10 tok/s, while-1,024 MiB fits all 12,288 with 2 GB of VRAM free.+A native (IQ) pack takes `--spec` 2 at the smallest: the engine refuses `--spec 0` and `--spec 1` on one (it needs `--native SHARD1`, `--spec T`+with T >= 2 and `--prefill CHUNK`), so speculative decoding cannot be turned off for these packs.++With `--stream-experts` under `STRATA_VERIFY_NO_HOST=1` (the `strata-sycl.sh` default) there is no SSD expert tier: every expert+must be in VRAM or in the pinned RAM mirror (`STRATA_MIRROR_MIB`), and the engine refuses to start when the two do not cover the model+("REFUSED: N experts are neither in VRAM nor mirrored"). Raise the mirror or lower `--max-context`. So the reserve matters: on the B70+with Coder IQ1_M 1,024 MiB fits all 12,288 experts with 2 GB of VRAM free.## Things that matter on this GPUdiff --git a/docs/INTEL_ARC.md b/docs/INTEL_ARC.mdindex 4adf07a3c..a38fab0d8 100644--- a/docs/INTEL_ARC.md+++ b/docs/INTEL_ARC.md@@ -47,6 +47,9 @@ engine package. Otherwise [build the image yourself](DOCKER.md#build-it-yourself`sudo apt install docker.io git`, then `sudo usermod -aG docker,render $USER` and log in again.- VRAM: the port keeps the experts on the card (`--stream-experts`). A 32 GB card holds the Coder IQ1_M or IQ2_XS. Smallercards mirror part of the experts in RAM and are slower.+ Under the wrapper's `STRATA_VERIFY_NO_HOST=1` there is no SSD expert tier on `xe`: every expert must+ be in VRAM or in the pinned RAM mirror (`STRATA_MIRROR_MIB`), and the engine refuses to start when+ the two do not cover the model. Size the RAM accordingly, or lower `--max-context`.## Build with Docker (what `setup --backend sycl` uses)