Skip to content

Releases: mindsdb/lightwood

Release 24.5.2.0

15 May 13:29
7ae7ea3
Compare
Choose a tag to compare

What's Changed

Full Changelog: v24.3.3.1...v24.5.2.0

Release 24.3.3.1

19 Mar 10:28
ec6171b
Compare
Choose a tag to compare

Changelog

Features

  • [ENH] Add test method to lightwood predictors #1206
  • [ENH] Weighted regression #1210

Fixes

  • [Fix] Check horizon in SkTime mixer #1202
  • [Fix] Manage and pin dependencies with Poetry #1209
  • [Fix] Remove unused dependencies #1212
  • [Fix] Doc build CI secret

Other

  • [Docs] Add documentation for Unit mixer #1204
  • [Fix] Updated docs link #1205
  • [Fix] Remove RnnEncoder #1203

Release 23.12.4.0

10 Jan 21:22
411343b
Compare
Choose a tag to compare

Changelog

Note: this release was overdue, the actual package version has been up in PyPI since December 25th, 2023.

Fixes

  • Fix extra requirements
  • Bump deps (#1201)

Release 23.11.1.0

10 Jan 21:19
be24227
Compare
Choose a tag to compare

Changelog

Note: this release was long overdue, the actual package version has been up in PyPI since October 26th, 2023.

Features

  • Improve documentation of device-related methods and docs regarding GPU usage #1185
  • New ts defaults & XGBoostArrayMixer #1191

Fixes

Other

  • Bump neuralforecast requirement #1195
  • Update statsforecast & sktime #1197

Release 23.8.1.0

07 Aug 15:55
9717e9e
Compare
Choose a tag to compare

Changelog

Features

  • Abort if there are no valid input features #1175
  • Additional checks for categorical TS #1176
  • Efficient TS predict #1179

Fixes

  • Filter encoder args in code from jsonai #1177

Other

  • Add support for Python 3.10 & 3.11 #1181
  • Separate json_ai/codegen modules #1183

Benchmarks

Release 23.7.1.0

03 Jul 19:08
2812838
Compare
Choose a tag to compare

Changelog

Enhancements & features

  • [ENH] Embedding mode #1163
  • [ENH] Improve transform_timeseries #1164
  • [ENH] Various improvements #1168
  • [ENH] Efficient nhits finetune #1171

Fixes

  • [FIX] Fix for transform_timeseries multiprocessing #1169

Benchmarks

Release 23.6.4.0

27 Jun 15:41
3262859
Compare
Choose a tag to compare

Changelog

This release features an improved time series pipeline which yields improved runtimes. Depending on the dataset size, this can be anywhere from 5x to > 100x.

Features

N/A

Fixes

  • [FIX] black version #1152
  • [FIX] More code formatter checks #1154
  • [FIX] Optimized datetime encoder #1155
  • [FIX] GluonTS mixer type error #1159
  • [ENH] Improved none checks, remove STLs #1161
  • [ENH] Vectorized __make_predictions in TS transforms #1162
  • [ENH] Faster TS transforms #1156

Other

  • [Refactor] Rename MP flag #1153

Release 23.6.2.0

14 Jun 00:30
51a3f03
Compare
Choose a tag to compare

Changelog

Optimizations for an average 7.5x speedup with little to no accuracy loss include:

v23.2.1.1

16 Mar 19:38
Compare
Choose a tag to compare

Hotfix:

  • Bound and update some dependencies.
  • Action to release on PyPI any GitHub release, not only when pushing to stable.

Release 23.3.2.0

14 Mar 22:42
68a71f8
Compare
Choose a tag to compare

Changelog

Features

  • TabTransformers mixer #1115
  • Faster gluon ts mixer inference #1110
  • Gluon static feats #1120

Fixes

  • Bump statsforecast==1.4.0 #1113
  • Bump pytorch to 1.13 #1114
  • Fix ICP softmax error #1117
  • Fixes for GluonTS trainer #1119
  • Bound pandas requirement #1122

Other

  • Improved quickstart example, Update README.md #1112

Benchmarks