0.4.1

0.4.1 (2024-12-12)

Fix

• tagging
• bump

[main f5742e5] bump: version 0.4.0 → 0.4.1
3 files changed, 10 insertions(+), 3 deletions(-)

Full Changelog: 0.4.0...0.4.1

0.4.0

0.4.0 (2024-12-12)

Feat

• adds workflow for pretraining a super-network (#173)
• allows to pass fabric (#162)

Fix

• release.yml: artifact upload for changelog (#211)
• release.yml: make bump dependent on test-code (#209)
• set cos sin in max_seq_len (#203)
• update readme (#193)
• release workflow (#202)
• fix dependency versions to avoid breaking CI (#195)
• colab tutorial notebook and typos (#188)
• addition and standardization of docstrings (#176)
• Remove rope_cache from max_seq_len setter (#181)
• Revert "ci: reusing unit-test.yml in release.yml" (#187)
• release workflow clash with branch protection rules (#184)
• Revert "fix: reworking release.yml to avoid clashing with branch protection rules" (#182)
• reworking release.yml to avoid clashing with branch protection rules (#180)
• set input variables as required positional arguments (#172)
• update tokenizers in pyproject.toml (#168)
• type hints and docstring for mkdocs build --clean --strict (#167)
• forcing deepspeed to use CPU for profiling FLOPS (#154)
• deprecate flexible mlp heads (#160)

[main 8d6acdc] bump: version 0.3.0 → 0.4.0
3 files changed, 31 insertions(+), 3 deletions(-)

What's Changed

• fix: deprecate flexible mlp heads by @rheasukthanker in #160
• fix: forcing deepspeed to use CPU for profiling FLOPS by @timurcarstensen in #154
• fix: fixed type hints by @zeqri in #167
• fix: update tokenizers in pyproject.toml by @rheasukthanker in #168
• feat: allows to pass fabric by @aaronkl in #162
• fix: set input variables as required positional arguments by @zeqri in #172
• ci: add python 3.12 support by @snikhil24 in #174
• fix: reworking release.yml to avoid clashing with branch protection rules by @timurcarstensen in #180
• fix: Revert "fix: reworking release.yml to avoid clashing with branch protection rules" by @timurcarstensen in #182
• fix: release workflow clash with branch protection rules by @timurcarstensen in #184
• ci: reusing unit-test.yml in release.yml by @timurcarstensen in #186
• fix: Revert "ci: reusing unit-test.yml in release.yml" by @timurcarstensen in #187
• feat: adds workflow for pretraining a super-network by @aaronkl in #173
• fix: Remove rope_cache from max_seq_len setter by @gabikadlecova in #181
• fix: addition and standardization of docstrings by @zeqri in #176
• fix: colab tutorial notebook and typos by @HannanJaved in #188
• fix: fix dependency versions to avoid breaking CI by @aaronkl in #195
• fix: release workflow by @timurcarstensen in #202
• fix: update readme by @aaronkl in #193
• fix: set cos sin in max_seq_len by @rheasukthanker in #203
• fix(release.yml): make bump dependent on test-code by @timurcarstensen in #209
• fix(release.yml): artifact upload for changelog by @timurcarstensen in #211

New Contributors

• @zeqri made their first contribution in #167
• @snikhil24 made their first contribution in #174
• @HannanJaved made their first contribution in #188

Full Changelog: 0.3.0...0.4.0

0.3.0

0.3.0 (2024-10-24)

Feat

• add support for LLamaMLP in extract_sub_network (#147)
• adding flops and macs profiling for subnets (#145)
• add script to profile latency (#141)
• modify rope for llama-3 and support llama-3.2 (#131)
• add gpt tutorial notebook utils (#122)
• add installation instruction to documentation (#121)

Fix

• refactor names of metric (#152)
• Extract weights for norm layers, test with random initialization. (#151)
• handle device in GPT model properly (#143)
• rename call function (#144)
• delete supernet_configs directory (#140)
• deprecate sample_random_indices (#133)
• support params, mag when sharing layer norm in phi-2 (#127)
• reset random layers in reset_super_network (#126)
• support GQA param count (#124)
• update readme (#111)

[main 41505ec] bump: version 0.2.0 → 0.3.0
3 files changed, 27 insertions(+), 3 deletions(-)

What's Changed

• fix: update readme by @aaronkl in #111
• feat: add installation instruction to documentation by @aaronkl in #121
• feat: add gpt tutorial notebook utils by @rheasukthanker in #122
• fix: support GQA param count by @rheasukthanker in #124
• fix: reset random layers in reset_super_network by @rheasukthanker in #126
• fix: support params, mag when sharing layer norm in phi-2 by @rheasukthanker in #127
• feat: modify rope for llama-3 and support llama-3.2 by @rheasukthanker in #131
• chore: renaming sinc to sine_curves by @timurcarstensen in #135
• fix: deprecate sample_random_indices by @rheasukthanker in #133
• fix: delete supernet_configs directory by @rheasukthanker in #140
• feat: add script to profile latency by @rheasukthanker in #141
• fix: rename call function by @aaronkl in #144
• ci: switching to uv for github actions by @timurcarstensen in #149
• feat: adding flops and macs profiling for subnets by @timurcarstensen in #145
• fix: handle device in GPT model properly by @aaronkl in #143
• feat: add support for LLamaMLP in extract_sub_network by @gabikadlecova in #147
• fix: Extract weights for norm layers, test with random initialization. by @gabikadlecova in #151
• fix: refactor names of metric by @aaronkl in #152

New Contributors

• @gabikadlecova made their first contribution in #147

Full Changelog: 0.2.0...0.3.0

0.2.0

0.2.0 (2024-09-08)

Feat

• litgpt update (#95)

Fix

• adding cz config (#119)
• removing version parsing in whittle/init.py (#118)
• renaming whittle/version to whittle/version.py (#117)
• commitizen configuration (#115)
• remove deprecated module (#105)
• delete old code (#104)
• allow to pass other loss function to training strategies (#101)
• set random state properly in sampler (#103)

[main e8f66d1] bump: version 0.1.3 → 0.2.0
3 files changed, 19 insertions(+), 3 deletions(-)
create mode 100644 CHANGELOG.md

What's Changed

• feat: litgpt update by @rheasukthanker in #95
• fix: set random state properly in sampler by @rheasukthanker in #103
• fix: allow to pass other loss function to training strategies by @aaronkl in #101
• fix: delete old code by @aaronkl in #104
• fix: remove deprecated module by @aaronkl in #105
• fix random sampler by @rheasukthanker in #110
• ci: automatic release workflow by @timurcarstensen in #107
• Revert "ci: automatic release workflow (#107)" by @timurcarstensen in #112
• ci: automatic release workflow by @timurcarstensen in #113
• fix: commitizen configuration by @timurcarstensen in #115
• Revert "fix: commitizen configuration" by @timurcarstensen in #116
• fix: renaming whittle/version to whittle/version.py by @timurcarstensen in #117
• fix: removing version parsing in whittle/init.py by @timurcarstensen in #118
• fix: adding cz config by @timurcarstensen in #119

Full Changelog: 0.1.3...0.2.0

v0.1.3

First working version of whittle