[None][fix] Respect KVCM V2 initialization and warmup budgets - #19213

Open

yizhang-nv wants to merge 8 commits into

NVIDIA:mainfrom

yizhang-nv:codex/fix-kvcm-v2-init-warmup-budget

Commits on Sep 17, 2026

[None][fix] Respect KVCM V2 initialization and warmup budgets
yizhang-nv
committed
5bb3aa8
[None][test] Preserve full and mixed attention constraint coverage
yizhang-nv
committed
49abcc6
[None][fix] Limit initialization floor changes to self attention
yizhang-nv
committed
10fef55
[None][test] Cover multimodal examples with KVCM V2 budget fix
yizhang-nv
committed
b595aec
[None][fix] Estimate V2 cache constraints and query warmup capacity
yizhang-nv
committed
f99fae5
[None][fix] Keep V2 budget fix focused on existing CI failures
yizhang-nv
committed
4c87a8d
[None][fix] Restore warmup capacity query after short request allocation
yizhang-nv
committed
9c3c53b
[None][test] Cover insufficient CUDA graph warmup capacity
yizhang-nv
committed
40466ac