v1.1.0

New Features

• Implement method for accessing residue-level representations by @yutanagano in #30

Other Changes

• Fix bug in docs landing page figure legend text by @yutanagano in #26
• Update branding by @yutanagano in #27
• Update README.md by @yutanagano in #28
• Improve docs by @yutanagano in #31
• Simplify docstrings by @andim in #32
• Clarify single chain models in docs by @yutanagano in #33

New Contributors

• @andim made their first contribution in #32

Full Changelog: v1.0.1...v1.1.0

v1.1.0-beta.1

What's Changed

• Fix bug in docs landing page figure legend text by @yutanagano in #26
• Update branding by @yutanagano in #27
• Update README.md by @yutanagano in #28
• Bump version by @yutanagano in #29
• Implement method for accessing residue-level representations by @yutanagano in #30
• Improve docs by @yutanagano in #31

Full Changelog: v1.0.1...v1.1.0-beta.1

v1.0.1

Minor patch to add links to the preprint on arxiv.org to the README and documentation.

v1.0.0

First production release! 🥳 Thank you to everyone who has helped with the project.

What's Changed

• Add latest model versions by @yutanagano in #19
• Make docs by @yutanagano in #20
• Improve docs by @yutanagano in #21
• Move functional api to top by @yutanagano in #22
• Add publication figures by @yutanagano in #23
• Prepare for first stable release by @yutanagano in #24

Full Changelog: v1.0.0-beta.1...v1.0.0

v1.0.0-beta.1

What's Changed

• Rename classic variant to left-aligned by @yutanagano in #11
• Update SCEPTR name acronym from ...Peptide... to ...Primary... by @yutanagano in #12
• Polish documentation by @yutanagano in #13
• Bump libtcrlm dependency to v1.0.0 by @yutanagano in #14
• Polish pyproject TOML by @yutanagano in #15
• Add github workflows by @yutanagano in #16
• Prepare for first beta release by @yutanagano in #17

Full Changelog: v1.0.0-alpha.4...v1.0.0-beta.1

v1.0.0-alpha.4

What's Changed

• Add finetuned model by @yutanagano in #4
• Rename models by @yutanagano in #5
• Add new OLGA variant by @yutanagano in #6
• Add more variants by @yutanagano in #7
• Add unpaired variant by @yutanagano in #8
• Add dropout noise only variant by @yutanagano in #9
• rename variants according to changes in paper draft by @yutanagano in #10

Full Changelog: v1.0.0-alpha.3...v1.0.0-alpha.4

v1.0.0-alpha.3

What's Changed

• ⚠️ BREAKING CHANGE: the xlarge variant is no longer tracked and published as it was unnecessarily large, unwieldy, and was no better than its smaller counterparts making it not a very interesting variant
• New variants added (classic variants, they use learnable initial token embeddings and a traditional sinusoidal positional information embedding method)

Full Changelog: v1.0.0-alpha.2...v1.0.0-alpha.3

v1.0.0-alpha.2

What's Changed

• Add new variants by @yutanagano in #2
• Track libtcrlm dependency version with explicit git commit hash to mitigate pip version conflict resolution issues

Full Changelog: v1.0.0-alpha.1...v1.0.0-alpha.2

v1.0.0-alpha.1

What's Changed

• Library code is now outsourced to a separate private repository libtcrlm. This means installing the newest version of sceptr requires access to this second private repo. If you would like access, please notify @yutanagano.
• New variants
  • single chain models
  • larger paired-chain variant
  • extra large CDR3-only, MLM-only variant
  • BLOSUM embedding variant (the initial embedding to the transformer uses BLOSUM amino acid encodings)

Full Changelog: https://github.com/yutanagano/sceptr/commits/v1.0.0-alpha.1