[Mooncake] Fix mixed MLA+Eagle block-size validation #39596
ywang96
merged
9 commits into
vllm-project:main
from
zhewenl:fix/mooncake-mla-eagle-assertion
Fix Mooncake MLA Eagle cache registration assertion
a5d5dd2a
gemini-code-assist
commented
on 2026-04-12
Sync Mooncake block size with kernel backends
a7e671d9
zhewenl
changed the title
test
54aa33fd
Merge branch 'main' into fix/mooncake-mla-eagle-assertion
86d18b16
chatgpt-codex-connector
commented
on 2026-04-12
NickLucche
approved these changes
on 2026-04-14
![mergify[bot]](https://avatars.githubusercontent.com/in/10562?v=4&s=60)
Merge branch 'main' into fix/mooncake-mla-eagle-assertion
f84fd121
[Bugfix] Fix MLA+Eagle test to set kv_topo.is_mla for correct split_k…
0c929981
disabled auto-merge 110 days ago
Head branch was pushed to by a user without write access
Head branch was pushed to by a user without write access
ywang96
commented
on 2026-04-14
Add comment explaining _sync_block_size_with_kernel is for Eagle alig…
f7169a7e
Merge branch 'main' into fix/mooncake-mla-eagle-assertion
5815b664
Merge branch 'main' into fix/mooncake-mla-eagle-assertion
3fbb0b9a
Assignees
No one assigned
Labels
ready
v1
kv-connector
Milestone
No milestone
Login to write a write a comment.
Login via GitHub