PR #16063 model : add BailingMoeV2 support

model : add BailingMoeV2 support #16063

CISC merged 22 commits into master from cisc/bailingmoe2

add BailingMoeV2 support

ece59aca

update llm types

d3392b0f

undo

ed6a6095

undo

dd3a80b8

update llm types

23f56932

add model collection link

35976bab

update

15ed1c66

CISC requested a review from

ggerganov 52 days ago

github-actions added python

ggerganov commented on 2025-09-18

CISC marked this pull request as draft 52 days ago

Merge branch 'master' into cisc/bailingmoe2

2ab54098

almost working

6ee701d9

CISC commented on 2025-09-27

Merge branch 'master' into cisc/bailingmoe2

a3904c0c

CISC commented on 2025-09-27

CISC marked this pull request as ready for review 43 days ago

CISC requested a review from

ngxson 43 days ago

CISC requested a review from

ggerganov 43 days ago

ggerganov commented on 2025-09-28

CISC commented on 2025-09-28

correct group selection and rename n_group_exp

f885540a

avoid large top_k and use argmax instead for now

39e80738

poke

4ea8a45f

CISC requested a review from

slaren 41 days ago

Merge branch 'master' into cisc/bailingmoe2

6870b160

Merge branch 'master' into cisc/bailingmoe2

362710a5

Merge branch 'master' into cisc/bailingmoe2

f82f0502

skip group selection when there are no tokens

e5c32f28

fix 1T conversion

950e9647

hopefully fixed expert group selection

bc6c48af

CISC requested a review from

ggerganov 23 days ago

make expert group selection generally available

6dd223bb

allow n_expert_groups to be 1 (Kimi K2) [no ci]

ec5fd81c

ggerganov approved these changes on 2025-10-20

address review suggestions

0e24c4ab

CISC merged 84bf3c67 into master 20 days ago

CISC deleted the cisc/bailingmoe2 branch 20 days ago

Reviewers

ggerganov

foldl

ngxson

slaren

Assignees

No one assigned

Labels

python

Milestone

No milestone

llama.cpp model : add BailingMoeV2 support #16063 Merged

model : add BailingMoeV2 support #16063

llama.cpp
model : add BailingMoeV2 support
#16063

Merged