Commits

Commits on Oct 17, 2024

Update router.py (#158 )
mihir-db
authored

Commits on Sep 19, 2024

Initialize default device lazily (#152 )
janEbert
authored

Commits on Sep 9, 2024

Implement Router Z-loss (#151 )
josejg
authored

Commits on Sep 5, 2024

Remove deprecated torch.cuda.amp custom fwd and bwd (#150 )
snarayan21
authored

Commits on Aug 31, 2024

bump _version.py to 0.7.0.dev0 (#148 )

eitanturok
and
snarayan21
authored
yo (#149 )
snarayan21
authored

Commits on Aug 30, 2024

Bump torch to <2.4.1 (#145 )

eitanturok
and
snarayan21
authored

Commits on Aug 28, 2024

Type Checking (#141 )
eitanturok
authored

Commits on Aug 12, 2024

Commits on Aug 9, 2024

Commits on Aug 5, 2024

only run GA if repo owner is Databricks (#135 )
eitanturok
authored

Commits on Aug 1, 2024

Clean up setup.py (#128 )
eitanturok
authored

Commits on Jul 31, 2024

Commits on Jul 17, 2024

Commits on Jul 11, 2024

dont save moe lb-loss tensors if args.moe_loss_weight=0 (#119 )

michael-go
and
Michael Gokhman
authored

Commits on May 21, 2024

Commits on May 17, 2024

Add Shared Expert (#109 )
vchiley
authored

Commits on May 15, 2024

Fix for ffn_hidden_size of 128, and better error message for incompatible ffn sizes. (#108 )
snarayan21
authored

Commits on Mar 26, 2024

Update README.md (#98 )
dakinggg
authored

Commits on Feb 20, 2024

Remove turbo (#96 )
dblalock
authored

Commits on Jan 11, 2024

Updt triton pin (#89 )
vchiley
authored

Commits on Jan 10, 2024

Fix moe_normalize_expert_weights when top_k=1 (#87 )
152334H
authored

Commits on Dec 22, 2023

fix default (#76 )
mvpatel2000
authored
refactor (#75 )
j316chuck
authored
Mem opt glu bkwd (#72 )

mvpatel2000
and
vchiley
authored