Releases: kakao/buffalo
Releases · kakao/buffalo
v2.0.5
What's Changed
- Update build_and_publush.yaml by @ita9naiwa in #83
- Bump version by @ita9naiwa in #85
Full Changelog: v2.0.4...v2.0.5
v2.0.4
What's Changed
- Add eALS by @dkkim1005 in #77
- Fix imports by @dkkim1005 in #79
- Fix type mismatch error in MatrixMarket by @dkkim1005 in #81
- Fix W2V stream data newline error. by @yupyub in #82
New Contributors
- @dkkim1005 made their first contribution in #77
- @yupyub made their first contribution in #82
Full Changelog: v2.0.3...v2.0.4
Release v2.0.3
What's Changed
- Fix 3rd by @lsh918 in #70
- Update by applying CLA Assistant by @oss-kakao in #71
- Temp file delete by @ita9naiwa in #73
- bug fix: progress bar in log.py by @pakhy2380 in #74
- fix bug during w2v training with utf8 characters by @pakhy2380 in #76
- bump version by @ita9naiwa in #78
Release v2.0.2
Changes
Bug fix
ALSOption
allows float typealpha
not aint
type.- Fixed CUDA compilation bug
- Changed module name (
aux
->_aux
) to avoid conflict with windows OS's reserved keywordfrom aurochs.misc.aux import xxx
would not work anymore- Instead, use
from aurochs.misc import aux; aux.xxx
orfrom aurochs.misc._aux import xxx
Misc
- Changed build backed (poetry -> setuptools)
- Changed submodule path for Eigen and its version(to 3.4)
- Removed unnecessary imports and directives from cython sources
- Removed useless dependency and CLI
fire
&absl-py
- Add Github acntions CI pipleline for building wheels
- Only supports Linux & MacOS wheel with
x86_64
architecture
- Only supports Linux & MacOS wheel with
Buffalo v2.0.1
- Build wheels on pypi are fixed
Buffalo v2.0.0
-
Added new algorithm WARP (Collaborative Metric Learning)
-
Add new training algorithm ials++ for learning ALS Model
-
Delete tensorboard, hyperopt
- Optimizable is replaced now
train
methods of models now taketraining_callback
e.g.) https://github.com/kakao/buffalo/blob/54c97fd300c2615eb53fcfd4fdc7879b98f19703/buffalo/algo/als.py#L144
-
Refactor import structure
- Now modules can be imported with shorter depth
-
previous
from buffalo.cfr import CFR
from buffalo.options import CFROption
from buffalo.data StreamOptions
from buffalo.misc import aux, log
- new
from buffalo import CFR, CFROption, StreamOptions
from buffalo import aux, log
- Now able to install via bdist wheel for python 3.8, 3.9, 3.10 on manylinux machines
Buffalo v1.2.2
-
Added new algorithm:
Probabilistic Latent Semantic Indexing(pLSI)
- Support warm-start
-
Delete existing temporary data file when creating new data file
-
Removed tensorflow dependency
-
Remove python n2 dependency
- Still needs in testing
- Resolved build time dependency error
-
Calculate idcg indpendently of rec result
-
Fix bug of W2V
- OOV caused by
min_cut
- OOV caused by
Buffalo v1.2.1
Updated
- Improved the speed of database creation.
Bug Fixes
- Fixed warp accuracy test code
- Fix typos
Release 1.2.0
Update
- Added new algorithm: Weighted Approximate-Rank Pairwise Matrix Factorization
- Weston, Jason, Samy Bengio, and Nicolas Usunier. "Wsabie: Scaling up to large vocabulary image annotation." Twenty-Second International Joint Conference on Artificial Intelligence. 2011.
- Check-out some experiments, and unit test code
- Added accuracy benchmarking results (work in progress)
Release 1.1.2
Bug fixes
- fix ImportError for 3rd/n2