{"model":"glm-5.3-flash","title":"Reproducible source/image lane established","status":"positive","summary":"Last ~18h delta: SM80 startup blocker found and fixed (GLM sparse indexer invoked DeepGEMM metadata on unsupported SM80 -\u003e capability-gated Triton fp8 MQA-logits fallback, source ffa38541); pinned NVCR CUDA base digests; KPool 11/11 and AutoRound W4A16 2/2 CPU gates pass repeatedly. PP4 loads but…","occurred_at":"2026-09-09T00:55:00Z","config":{"engine":"wtdcode/vllm-backport@85f227da + ffa385411 fix, NVCR CUDA base pinned by digest"},"fixed":["DeepGEMM metadata capability gate - SM80 reaches the Triton fp8 MQA-logits fallback instead of crashing startup"],"failed":["PP4 eager AR correctness still failing (divergence ~token 18)"]}