[None][chore] BREAKING: Test KV cache manager v2 as the default - #17495

Draft

yizhang-nv wants to merge 16 commits into

NVIDIA:mainfrom

yizhang-nv:codex/kvcm-v2-default-true-ci

Commits on Oct 1, 2026

[None][feat] Default to KVCM V2 and Python transceiver
yizhang-nv
committed
f926028
[None][test] Skip known KVCM V2 failures
yizhang-nv
committed
edcd8dd
[None][test] Skip latest KVCM V2 failures
yizhang-nv
committed
b306dc4
[None][test] Skip concurrent KVCM V2 OOM
yizhang-nv
committed
0122bf9
[None][test] Re-enable fixed and rerunnable KVCM V2 tests
yizhang-nv
committed
5706010
test: remove redundant V1 disaggregation coverage
yizhang-nv
committed
4054c9a
[None][fix] Use default KV cache manager for aggregated DeepSeek V3.2
yizhang-nv
committed
33e3630
[None][test] Pin v1 disagg metrics and update disagg coverage notes
yizhang-nv
committed
6074278
[None][chore] Remove local disagg review documents
yizhang-nv
committed
e558bc2
[None][test] Waive single-GPU failures pending KVCM V2 fixes
yizhang-nv
committed
976b6ec
[None][test] Remove stale V1 perf overrides and redundant test
yizhang-nv
committed
3449e57
[None][test] Re-enable Qwen3 LoRA after merged V2 test fix
yizhang-nv
committed
d7bc57e
[None][test] Re-enable dynamic draft after merged warmup fix
yizhang-nv
committed
525cafe
[None][test] Update KVCM V2 CI coverage and waivers
yizhang-nv
committed
330c05a
[None][test] Unwaive KVCM V2 cases except initialization OOM
yizhang-nv
committed
b6c19ed
[None][test] Initialize max beam width in KV cache budget fixture
yizhang-nv
committed
97c80d2