Skip to content

Latest commit

 

History

History
221 lines (102 loc) · 8.87 KB

CHANGELOG.md

File metadata and controls

221 lines (102 loc) · 8.87 KB
Raw

CHANGELOG

v0.6.2 (2024-08-07)

Chore

  • chore: update repository / homepage metadata (b917d47)

Fix

  • fix: bug with gemma cache on non-fp32 (#10) (52a9fb7)

v0.6.1 (2024-08-07)

Fix

  • fix: fix cache behavior for gemma2 and warn if caching breaks (#9) (18a7bfc)

v0.6.0 (2024-08-04)

Feature

  • feat: adding LRE forward() method (#8) (9558e5f)

v0.5.2 (2024-08-04)

Fix

  • fix: get_device should gracefully handle models without a device attr (#7) (5fb2681)

v0.5.1 (2024-08-04)

Fix

  • fix: relax transformers version (fd7b86c)

v0.5.0 (2024-06-30)

Feature

  • feat: improving Lre repr() (011d936)

v0.4.0 (2024-06-29)

Feature

  • feat: add map_activations_fn param in ConceptMatcher (#6)

  • feat: allow passing a map_activations_fn to ConceptMatcher

  • it seems tranformers >4.42.0 is broken (1ac957a)

v0.3.1 (2024-05-02)

Fix

  • fix: fixing bug in inverted LRE forward() (46510ae)

v0.3.0 (2024-05-02)

Chore

  • chore: adding codecov (#4) (d25bc48)

  • chore: Update README.md (d35bc8b)

  • chore: removing unused constant (5b60e5b)

Feature

  • feat: allow setting max prompts in LRE training (#5) (2066999)

v0.2.1 (2024-02-16)

Fix

  • fix: exporting ObjectAggregation and VectorAggregation from main module (975e215)

v0.2.0 (2024-02-16)

Chore

  • chore: update setup-python action to use node 20 (18051cc)

Feature

  • feat: allow passing a callable as inv_lre to train_relation_concepts_from_inv_lre() (#2) (12fa48e)

v0.1.4 (2024-02-15)

Chore

  • chore: fixing typing for newer torch types (3f668d9)

  • chore: fix typo in ci.yaml (2492912)

  • chore: fixing typo in docs (a6ddd0b)

Fix

  • fix: adding py.typed marker to declare type inference (5249dd3)

v0.1.3 (2023-11-23)

Fix

  • fix: match dtype of activations during Concept.forward (3a06459)

v0.1.2 (2023-11-23)

Chore

  • chore: Update README.md (9e85cf0)

  • chore: improving docs (c1fb709)

  • chore: improving docs (ab4748e)

  • chore: updating docs with advanced usage (7439e71)

  • chore: Add docs link to README.md (b915827)

  • chore: typo in docs (b4f547a)

Fix

  • fix: set device when checking if answers match (008c4b6)

Unknown

v0.1.1 (2023-11-23)

Chore

  • chore: add Sphinx docs (#1)

  • adding a basic docs page and validating in ci

  • adding deploy pages action (4088d61)

  • chore: Adding badges to README.md (341b6cd)

Fix

v0.1.0 (2023-11-22)

Feature

v0.0.0 (2023-11-22)

Unknown

  • moving protobuf to dev deps (a6fdcec)

  • moving sentencepiece to dev reqs (03e38f5)

  • fixing python version (bf451de)

  • synchronizing version (0bc4130)

  • setting up deployment (369660c)

  • minor change to readme (9dc7a81)

  • removing name of paper from links (8bd04f0)

  • renaming batch params to be more specific (b6c70b4)

  • removing caching as it throws errors (53649e3)

  • updating dependency cache (4ff0098)

  • adding back dataclasses-json dep (30eef6a)

  • removing poetry.lock (5237bfb)

  • removing unused dependency (305ed04)

  • Update README.md (6839ca0)

  • tweaking README and adding base module exports (902ee43)

  • updating README (1bd8e31)

  • improving tests for trainer (c9d0f69)

  • default to guessing layer matcher if not provided (9c52625)

  • adding a helper to guess the hidden layers of the model (c1c01d6)

  • adding a trainer class (fcf673d)

  • adding citation info (787b6de)

  • adding initial libs and classes (2691e8f)

  • Initial commit (2ccf587)