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++
gyro_setup.sh: point the HIP build at ROCm (ROCM_PATH or /opt/rocm)
Laurent Zuijdwijk2fce1e1e6156
diff --git a/tools/gyro_setup.sh b/tools/gyro_setup.shindex 2d3899e9f..3282c8cdd 100755--- a/tools/gyro_setup.sh+++ b/tools/gyro_setup.sh@@ -35,7 +35,8 @@ if [ "$BACKEND" = cuda ]; thenelif [ "$BACKEND" = hip ]; thenecho "note: AMD is experimental for Gyro (kernel parity validated on gfx1151, no end-to-end run yet)"cmake -S . -B build -G Ninja -DCMAKE_BUILD_TYPE=Release -DSTRATA_ENABLE_HIP=ON -DSTRATA_PREFILL_MMQ=ON \- -DCMAKE_HIP_ARCHITECTURES="$HIP_ARCH" -DSTRATA_GGML_DIR="$LLAMA"+ -DCMAKE_HIP_ARCHITECTURES="$HIP_ARCH" -DSTRATA_GGML_DIR="$LLAMA" \+ -DCMAKE_PREFIX_PATH="${ROCM_PATH:-/opt/rocm}"else echo "--backend must be cuda or hip"; exit 2; ficmake --build build -j