transformers
10f60888
- Merge branch 'main' into load-convert-before-shard
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
9 days ago
Merge branch 'main' into load-convert-before-shard
References
#48204 - [Umbrella] FSDP2 + expert-parallel token dispatch, with sharded-loading prefetch
#48519 - Loading: convert transposed and RoPE-permuted weights before taking the DTensor shard
Author
qgallouedec
Parents
5fa9a03d
606e6e8b
Loading