Fix EP post merge #43730

ArthurZucker merged 35 commits into main from fix-ep
ArthurZucker
ArthurZucker ArthurZucker force pushed from 629d3038 to c084aa7a 153 days ago
ArthurZucker restore
ec2ed1db
ArthurZucker ArthurZucker force pushed from 596897f7 to ec2ed1db 153 days ago
ArthurZucker add all reduce for ep
33ca3308
HuggingFaceDocBuilderDev
ArthurZucker fix init and bias sharding
e545ac1f
ArthurZucker fix finalize weight init
fa78068d
ArthurZucker add full stacktracing
6e4d2343
ArthurZucker ArthurZucker changed the title up Fix EP post merge 152 days ago
ArthurZucker fix
05fc1fa2
ArthurZucker okay big improvement here
819698c6
ArthurZucker the only case shard index should be used is when we are acctually col…
d99f8340
3outeille
3outeille commented on 2026-02-04
ArthurZucker more fixes
f0d0de12
ArthurZucker fix EP forward gpt oss
c5cbdc80
ArthurZucker revert some shit
94d676cd
ArthurZucker when you are stupid sometimes you really need a brain :) :) :) :)
959b46f7
ArthurZucker fix TP
01c5774d
ArthurZucker Ok GPT oss is fixed now
9dbb634c
ArthurZucker try to fix perms
83742984
ArthurZucker attempt to fix
8e46655c
ArthurZucker am I a doomer and AI is not that bad?
104f80d3
ArthurZucker fix
14dca0c1
ArthurZucker it "passes" but the output is shit
3600fbe4
ArthurZucker Merge branch 'fix-ep' of github.com:huggingface/transformers into fix-ep
fabde8a2
ArthurZucker style my man
20dee9a8
ArthurZucker outputs are gonna be giberish but at least the forward pass "works"
0b95c643
ArthurZucker Merge branch 'main' of github.com:huggingface/transformers into fix-ep
9137fa25
ArthurZucker dtyle
646cbe38
ArthurZucker fix mixtral
4b32a6b2
ArthurZucker okay shape fixes
14607626
ArthurZucker tensor idx is only for groupped gemm / EP
f7893956
ArthurZucker fix gate_up shard
8b4ed7bd
ArthurZucker fix :)
d8cd5339
ArthurZucker revert some EP changes that are breaking other stuff
76c904cf
ArthurZucker style
cfd92d79
3outeille
3outeille commented on 2026-02-05
3outeille
3outeille commented on 2026-02-05
3outeille
3outeille commented on 2026-02-05
3outeille
3outeille commented on 2026-02-05
3outeille use moe_tp_experts
7ad14a70
github-actions
3outeille
3outeille
3outeille approved these changes on 2026-02-05
ArthurZucker revert unrelated, last nits and style
c3ae670f
ArthurZucker good?
45dd1248
ArthurZucker fix modlar
46900e07
ArthurZucker ArthurZucker merged e1f37667 into main 151 days ago
ArthurZucker ArthurZucker deleted the fix-ep branch 151 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone