v0.0.6

What's Changed

• Bump actions/checkout from 4 to 5 by @dependabot[bot] in #97
• Bump actions/upload-pages-artifact from 3 to 4 by @dependabot[bot] in #96
• Bump actions/download-artifact from 4 to 5 by @dependabot[bot] in #95
• Update pytest-cov requirement from <7,>=5 to >=5,<8 by @dependabot[bot] in #98
• Run actions on new pr by @fschlatt in #99
• Add LoRA Support by @lgienapp in #94
• Fix linters by @fschlatt in #100
• Remove complete HF cache after each test to save space in runner by @fschlatt in #101
• Bump actions/setup-python from 5 to 6 by @dependabot[bot] in #103
• Move rank-distillm to hf by @fschlatt in #104
• Rename normalize to normalization by @fschlatt in #105
• Cleanup run file loading by @fschlatt in #106
• External model hub improvments by @fschlatt in #107
• Fix test actions by @fschlatt in #108
• Fix mean nan mean aggregation by @fschlatt in #110
• Ignore errors on removing hf cache by @fschlatt in #111
• Add temperature to ranknet by @fschlatt in #112
• Add unicoil by @fschlatt in #113
• Add more splade models by @fschlatt in #114
• Fix variable name by @fschlatt in #115
• Fix mean and sum aggregation with masking by @fschlatt in #116
• fix trailing whitespace by @fschlatt in #117
• Add another opensearch splade model by @fschlatt in #118
• Fix lexical splade by @fschlatt in #119
• Fix dtypes for naive exact searchers by @fschlatt in #120
• Return loss and output on training step by @fschlatt in #121
• add option to manually configure backbone model by @fschlatt in #124
• fix linters by @fschlatt in #125
• use dashed docs dataset id for saving indexes by @fschlatt in #126
• Make first pooling work on m chip macs. by @hscells in #127
• Add PLAID indexer and searcher implementations by @RaykKretzschmar in #102
• Bump actions/upload-artifact from 4 to 5 by @dependabot[bot] in #128
• Bump actions/download-artifact from 5 to 6 by @dependabot[bot] in #129
• Update sphinx version by @fschlatt in #131
• Make query and doc length optional + cleanup (cross-encoder) tokenizer by @fschlatt in #132
• Update pytest requirement from ==8.*,>=8.0 to >=8,<10 by @dependabot[bot] in #134
• Add tests for external datasets and cleanup external datasets by @fschlatt in #135
• Add XTR model implementation and tests by @RaykKretzschmar in #136
• migrate everything to pyproject.toml by @fschlatt in #138
• Cleanup pooling, normalization, and sparsification by @fschlatt in #137
• Version bump by @fschlatt in #139

New Contributors

• @lgienapp made their first contribution in #94

Full Changelog: v0.0.5...v.0.0.6

v0.0.5

What's Changed

• Update action to manually deploy docs by @fschlatt in #28
• Remove work in progress sbert integration by @fschlatt in #30
• Fix documentation for custom datasets by @fschlatt in #31
• Update docs class template by @fschlatt in #32
• Minor documentation improvements by @fschlatt in #35
• add scores for NanoBEIR dataset by @RaykKretzschmar in #33
• Add prepare data method to download data in single process and prevent data corruption by @fschlatt in #36
• Fix preparing data by sampling the first element from each ir dataset by @fschlatt in #37
• Fix col model document expansion by @fschlatt in #38
• Batch similarity functions for dense and sparse retriever to avoid OOM for large indexes by @fschlatt in #39
• Version bump in CITATION.cff by @fschlatt in #40
• Fix preparing docs store for to avoid corrupted data by @fschlatt in #41
• Correctly use batching for sparse and dense PyTorch indexer to avoid memory issues by @fschlatt in #42
• Speed up batching by using empty instead of zeros by @fschlatt in #43
• Restructure model-zoo tables and improve documentation clarity by @RaykKretzschmar in #44
• Add callbacks for indexing and searching in model-zoo by @RaykKretzschmar in #45
• Improve printing results for searching and re-ranking by @fschlatt in #46
• Increase precision for sparse and dense searcher by @fschlatt in #47
• Do not print results for validation in training by @fschlatt in #48
• Free memory in searching + Fix skip dataset logging by @fschlatt in #49
• Add index and save directory configurations for callbacks by @RaykKretzschmar in #50
• Fix logging for validation by @fschlatt in #51
• Fix getting dataset for single dataloader by @fschlatt in #52
• Fix class references in quickstart guide for models and configurations by @RaykKretzschmar in #53
• Fix class references in documentation by @RaykKretzschmar in #54
• Fix emptying cache when loading new index by @fschlatt in #55
• Minor dataset improvements by @fschlatt in #56
• Add convenience function to parse sparse CSR tensor by @fschlatt in #57
• Add the option to specify an index name by @fschlatt in #58
• Minor indexing and retrieval improvements by @fschlatt in #59
• Add ModernColBERT support and update tokenizer warnings by @RaykKretzschmar in #60
• Fix CI + Minor pretty printing fix by @fschlatt in #61
• Fix CI by @fschlatt in #62
• Add option to pass model kwargs when loading model using module by @fschlatt in #63
• Fix pretty printing for skipped inference datasets by @fschlatt in #64
• Improve passing model kwargs and fix quotes in saving run files by @fschlatt in #65
• Add monoelectra rank-distillm data by @fschlatt in #66
• Remove flash + Make model registration public by @fschlatt in #67
• Update seismic test index by @fschlatt in #68
• Entirely remove indexes from test data by @fschlatt in #69
• Return none for dataset len if no length can be inferred by @fschlatt in #71
• Fix doc dataset length and update docstrings by @fschlatt in #72
• Support different tokenizers for adding marker tokens by @fschlatt in #73
• Force query and doc ids to str by @fschlatt in #74
• Make cross-encoder abstract by @fschlatt in #76
• Fix tests by @fschlatt in #77
• Fix flake8 issues in documentation strings by @fschlatt in #78
• Remove colbert-ai from test dependencies by @fschlatt in #79
• Make sure datasets>=3.0.0 by @fschlatt in #80
• Refactor documentation to Google-style format by @RaykKretzschmar in #81
• Improve readability and documentation of loss functions by @fschlatt in #83
• Refactor documentation by @RaykKretzschmar in #82
• Fix mono models for transformers version bump by @fschlatt in #84
• Add Coil to Lightning IR by @samiki-hub in #16
• Fix coil without gpu by @fschlatt in #89
• Update COIL embedding docstrings by @fschlatt in #90
• Add docstrings for retrieve modules by @RaykKretzschmar in #88
• Add MVR to lightning-IR by @lisalehna in #75
• Improve documentation by @fschlatt in #92
• Version bump by @fschlatt in #93

New Contributors

• @samiki-hub made their first contribution in #16
• @lisalehna made their first contribution in #75

Full Changelog: v0.0.4...v0.0.5

v0.0.4

What's Changed

• Fix Testing with New Dataset Loading by @fschlatt in #20
• Fix MLM Decoder Weigth Tieing When Tieing Query-Document Projection by @fschlatt in #21
• Improve auto factory for models with multiple architectures by @fschlatt in #22
• Make BiEncoderModel an abstract base class and require abstract encode method + Split bi-encoder into single-vector and multi-vector by @RaykKretzschmar in #23
• Update Seismic Indexer and Searcher by @fschlatt in #24
• Minor documentation improvements by @fschlatt in #25
• Minor linting fixes and bug fixes by @fschlatt in #26
• Minor action fixes and version bump by @fschlatt in #27

New Contributors

• @fschlatt made their first contribution in #20
• @RaykKretzschmar made their first contribution in #23

Full Changelog: v0.0.3...v0.0.4

0.0.3

What's Changed

• docs: update README.md by @eltociear in #3
• Configure Dependabot updates by @janheinrichmerker in #8
• Bump actions/setup-python from 4 to 5 by @dependabot in #10
• Update pytest-cov requirement from ==5.*,>=5.0 to >=5,<7 by @dependabot in #9

New Contributors

• @eltociear made their first contribution in #3
• @janheinrichmerker made their first contribution in #8
• @dependabot made their first contribution in #10

Full Changelog: v0.0.2...v0.0.3

0.0.2

Minor updates and tweaks

Initial Release

Initial Lightning IR release