What's Changed
- Add natten with static tiling & Morton Curve by @joydddd in #87
- Fix Ruff by @drisspg in #90
- Remove persisted credential by @drisspg in #91
- Add
tabulate
used in example to dependencies by @oraluben in #99 - Reduce memory usage by @oraluben in #100
- Fix typos:
casual
->causal
by @awgu in #102 - Add end-to-end example for paged attention by @BoyuanFeng in #104
- Adapt paged attention implementation from pytorch by @BoyuanFeng in #105
- Add option for causal by @drisspg in #107
- fix ci by @drisspg in #114
- Add example for showcasing how to do multi-latent Attention by @drisspg in #113
- Change import structure by @drisspg in #125
- Fix natten x y dims by @fcossio in #128
- add batchify mask-mod by @drisspg in #132
- Make matplotlib lazy by @drisspg in #134
- Proper doc string deepseek ref in latent_attention.py by @curlup in #139
- Add Sliding Tile Attention impl by @yibozhong in #140
- Update workflows by @drisspg in #141
- Fix Publish by @drisspg in #142
New Contributors
- @joydddd made their first contribution in #87
- @oraluben made their first contribution in #99
- @awgu made their first contribution in #102
- @BoyuanFeng made their first contribution in #104
- @fcossio made their first contribution in #128
- @curlup made their first contribution in #139
- @yibozhong made their first contribution in #140
Full Changelog: v0.0.3...v0.0.4