Retune for L40S (48GB): CUDA graphs back on, roomier memory split 3469a45
nxj18 commited on
Commit History
Stability + persistence: OOM headroom, supervised vLLM, bucket cache/logs, open root ba215e5
nxj18 commited on
Install onnxruntime-gpu LAST (insightface drags in the CPU wheel) 8c70680
nxj18 commited on
Fix CUDA provider load for onnxruntime-gpu d543af1
nxj18 commited on
vLLM JoyCaption (OpenAI-compatible) + GPU InsightFace faces endpoint behind one authed gateway 939f281
nxj18 commited on