Run frontier MoE models on hardware you already own — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦
C
tests: the staged-import probe runs in staged_window (qwen-staged), whose job has the fixture dependencies
JustVugg37d8063ffb97
diff --git a/c/tests/vulkan_engines.sh b/c/tests/vulkan_engines.shindex ef0ddf62..23526534 100755--- a/c/tests/vulkan_engines.sh+++ b/c/tests/vulkan_engines.sh@@ -134,7 +134,6 @@ family_staged() {tail -1 vk_chain.log | grep -qx PASS || { cat vk_chain.log vk_chain.err; fail "staged chain ops"; }staged_check vk_chain.err "staged chain ops"echo "OK staged chain ops: $(grep -o 'resident data in host memory: .*' vk_chain.err)"- staged_import}# imported_copies <log>: the copies the exit report says came straight from host memoryimported_copies() { sed -n 's/.* MiB of them straight from host memory in \([0-9]*\) copies.*/\1/p' "$1" | tail -1; }@@ -188,6 +187,7 @@ staged_window() {chain_gate qwen36 "window qwen36 chain" 1 COLI_VK_HOST_VISIBLE_CAP_MB=246 COLI_DENSE_I8=0 SNAP=qwen36_tiny_c -- 8 8 qwen36_tiny/ref_full.jsonstaged_check vk.log "window qwen36 chain"echo "OK window qwen36: staged on its own, $(grep -o 'resident data in host memory: .*' vk.log)"+ staged_import # here: the engine's fixture needs the fixture dependencies, which staged's job has not}# Staged uploads failing (COLI_VK_STAGED_FAULT=<point>[:n], the n-th time) at every point