club-3090

Files

noonghunna ffbcb64dcf docs(multi-max): caveat — fp8 KV flip needs fresh 4-card re-validation

The prior 4-card validation (@Whamp #446) was on the OLD int8-PTH KV. #595 flips
to fp8/e4m3; the fp8 config is validated on the 2-card dual-max proxy (all gates
green) but not yet re-confirmed at TP=4. Request a fresh 4-card report before
upgrading ⚠️ caveats -> ✅.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EfF565T9eSLaqGzidyJ1Pm

2026-07-06 15:00:52 +00:00

agents-a1/vllm

DeepGEMM: cover fp8-dynamic weights too (agents-a1) — all fp8 slugs 5090-safe

2026-07-05 23:20:06 +00:00

diffusiongemma-26b-a4b/vllm

fp8 composes: add VLLM_USE_DEEP_GEMM pass-through parity + drift guard

2026-07-05 22:46:20 +00:00

gemma-4-12b

fold(gemma-12b): dual-bf16 onto vllm-stable v0.24.0, MTP-off

2026-07-01 12:13:25 +00:00

gemma-4-26b-a4b

fold(gemma-26b): MTP-off on v0.24.0 dual (Gemma-4 MTP×tools broken)