[None][fix] Unwaive supported attention backend cases - #15838

Merged

yuxianq merged 13 commits into

NVIDIA:mainfrom

yuxianq:unwaive-attn-backend-test-cases

Jul 8, 2026

Commits on Jul 7, 2026

[None][fix] enable TRTLLM MLA context backend tests
yuxianq
committed
bb93032
[None][fix] enable TRTLLM Blackwell MLA generation
yuxianq
committed
ccf8516
[None][fix] reduce TRTLLM attention backend exclusions
yuxianq
committed
3e2a315
[None][fix] unwaive stable attention backend cases
yuxianq
committed
ca66d0c
[None][fix] remove obsolete Hopper FMHA workaround
yuxianq
committed
4f259ca
[None][fix] support large FlashInfer KV appends
yuxianq
committed
52d1b18
[None][fix] restore missing MLA generation kernel guard
yuxianq
committed
96f9813
[None][fix] support FP8 KV MLA in TRTLLM-gen
yuxianq
committed
aadd5a4
[None][fix] fall back for FP8 multi-token MLA
yuxianq
committed
ec49210
[None][fix] gate FP8 multi-token MLA fallback by FlashInfer version
yuxianq
committed
f497c9c
[None][fix] remove obsolete FlashInfer MLA fallback
yuxianq
committed
a385754
[None][fix] fix FP8 KV-only attention fallback
yuxianq
committed
3a40842
[None][fix] avoid forcing paged context FMHA for FP8 KV
yuxianq
committed
7483843