Skip to content

Readme updates - #52

Merged
EiffL merged 4 commits into
mainfrom
readme_updates
Oct 21, 2025
Merged

Readme updates#52
EiffL merged 4 commits into
mainfrom
readme_updates

Conversation

@EiffL

Copy link
Copy Markdown
Collaborator

No description provided.

@EiffL
EiffL merged commit e9f661e into mainOct 21, 2025
2 checks passed
amina1315 added a commit to amina1315/AION that referenced this pull request Aug 15, 2026
* minor tweak
* update readme
* Update publish-pypi.yml
* Update pyproject.toml
* Adding more information on pypi page
* Update README.md
* Update README.md
* Update README.md
* updating readme slightly
* fix formatting issues
* Created using Colab
* improve documentation
* fixing formatting
* Update README.md
* Readme updates (PolymathicAI#52)
* improving Readme for release
* small update
* fix alignment
* simplify
* Readme updates (PolymathicAI#53)
* improving Readme for release
* small update
* fix alignment
* simplify
* adding acknowledgements
* Created using Colab
* updating notebook
* Readme updates (PolymathicAI#54)
* improving Readme for release
* small update
* fix alignment
* simplify
* adding acknowledgements
* adding arxiv link
* add spectrum preprocessing fns
* add pad length for SDSS and DESI spectra
* preprocessing in torch
* make padding operate on Spectrum class; apply padding in codec
* fix assignment
* fix edge case
* updating notebook (PolymathicAI#65)
* Notebook update (PolymathicAI#66)
* updating notebook
* updating dataset
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* Created using Colab
* fix(spectrum): use 'wavelength' key in pad_spectrum sentinels
`pad_spectrum` looks up sentinel values in `padding_values` while
iterating over the attribute names of `Spectrum` (`flux`, `ivar`,
`mask`, `wavelength`), but the dict key is `"lambda"`, so the
`wavelength` lookup misses and the wavelength tensor is padded with
the default `0` instead of the intended `99999`.
That makes the wavelength array non-monotonic
(`[real_min, …, real_max, 0, 0, …, 0]`). `LatentSpectralGrid.to_latent`
feeds it to `interp1d`, whose out-of-range mask is
`(xnew < x[..., 0]) | (xnew > x[..., -1])`. With `x[..., -1] == 0`,
every latent grid position satisfies `xnew > 0` and the whole latent
representation is replaced with `mask_value = 0.0`. Every spectrum then
encodes to (essentially) the same null embedding.
This shows up in downstream consumers that call the spectrum codec on
inputs whose length is less than `pad_length` (so `F.pad` actually
appends elements). Switching the dict key to `"wavelength"` so it
matches the attribute name fixes the sentinel.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Update README.md
---------
Co-authored-by: lhparker1 <liamholdenparker@gmail.com>
Co-authored-by: Francois Lanusse <EiffL@users.noreply.github.com>
Co-authored-by: Francois Lanusse <fr.eiffel@gmail.com>
Co-authored-by: Jeff Shen <jshen2014@hotmail.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: Tom Hehir <148493038+tom-hehir@users.noreply.github.com>
Sign up for freeto join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant

@EiffL