Skip to content

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++

tests: iq_multi_parity runs with STRATA_KQ_SPLIT=1

Niko12219c4516509f84
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 82249d4b3..fda44e24a 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -963,6 +963,7 @@ if(STRATA_ENABLE_CUDA OR STRATA_ENABLE_HIP)
target_include_directories(iq_multi_parity PRIVATE ${_strata_gpu_include_directories})
target_link_libraries(iq_multi_parity PRIVATE strata_kernels ${_strata_gpu_runtime_target})
add_test(NAME iq_multi_parity COMMAND iq_multi_parity)
+ set_tests_properties(iq_multi_parity PROPERTIES ENVIRONMENT "STRATA_KQ_SPLIT=1") # the Q4_K / Q5_K / Q8_0 Split path against the old kernels
# the prompt path's DeltaNet recurrence: gdn_rec_kh_kernel against gdn_rec_cols_pipe_kernel, bitwise (GPU,
# synthetic, no model; --bench adds timings). NVIDIA only: cp.async and the CUDA occupancy API (not on gfx906)
if(STRATA_ENABLE_CUDA AND NOT STRATA_HIP_GFX906)