[None][fix] Scope KVCM warmup capacity constraints to DeepSeek V4 - #19213

Open

yizhang-nv wants to merge 11 commits into

NVIDIA:mainfrom

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

Commits on Sep 20, 2026

[None][fix] Respect KVCM V2 initialization and warmup budgets
yizhang-nv
committed
ef0361a
[None][test] Preserve full and mixed attention constraint coverage
yizhang-nv
committed
10f389c
[None][fix] Limit initialization floor changes to self attention
yizhang-nv
committed
7a24133
[None][test] Cover multimodal examples with KVCM V2 budget fix
yizhang-nv
committed
d39cfb0
[None][fix] Estimate V2 cache constraints and query warmup capacity
yizhang-nv
committed
72bfd75
[None][fix] Keep V2 budget fix focused on existing CI failures
yizhang-nv
committed
8995290
[None][fix] Restore warmup capacity query after short request allocation
yizhang-nv
committed
51c69ff
[None][test] Cover insufficient CUDA graph warmup capacity
yizhang-nv
committed
3be8a74
[None][fix] Keep maximum-length warmup constraints specific to DeepSeek V4
yizhang-nv
committed
4767601
[None][test] Keep KVCM constraint test changes minimal
yizhang-nv
committed
28d08c0
[None][fix] Preserve generic context warmup constraints
yizhang-nv
committed
8b6febf