=== GPU RELEASE VERIFICATION 2026-08-16T16:33:17-04:00 ===
--- nvidia-smi ---
850 MiB, 32607 MiB, 1 %
--- GPU compute procs ---
3399, 42 MiB
3983, 72 MiB
15480, 85 MiB
69719, 79 MiB
--- llama-server processes ---
NONE
--- port 8195 ---
DEAD
--- port 8109 (qwen38 production, must be untouched) ---
not listening (as expected, unit disabled)
--- RAM ---
               total        used        free      shared  buff/cache   available
Mem:             188          27          27           0         146         160
--- vault free ---
/dev/nvme1n1p1  3.6T  3.1T  407G  89% <VAULT>
--- build dirs untouched ---
<VAULT>/models/llamacpp-b10453 <SIX OTHER MODEL BUILD DIRECTORIES ON THIS MACHINE, NAMES GENERALIZED> 
--- no llamacpp-muse created ---
correctly absent