add_axk1 #46867
vasqu
merged
19 commits into
huggingface:main
from
kmswin1:feature/add_axk1
add_axk1
fe898eb7
fix TP bug of load balancing bias term
89848e18
change default value of n_routed_experts and max_position_embeddings
b4f1355e
address review comments
af8e7e78
Merge remote-tracking branch 'upstream/main' into feature/add_axk1
f9b148a4
change model type to snake case, and rebase
e891d74c
refactor AXK1 to inherit from deepseek_v3 and modernize config
0f5550f7
address second review round: youtu-style tests, docs cleanup, inherit…
e14ffd51
use Expectations class and add batched logits integration test
2fb3bf47
Merge branch 'main' into feature/add_axk1
9d00b514
fixup repo
8e321a52
vasqu
approved these changes
on 2026-07-23
idk why but go
57bc5701
Add random small model test
b75ba4d0
test fixes
3e872cc5
oops that was from my local
eb020713
vasqu
commented
on 2026-07-23
vasqu
commented
on 2026-07-23
add integration test
87384de7
fixup integration tests
d375f5e7
Merge branch 'main' into feature/add_axk1
605dcf6c
style
da23893e
Login to write a write a comment.
Login via GitHub