Jax/PyTorch Backend #34
Open
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
scikit-learn
merge to latest stella
…ocs; add scripts; bump Python
…remove dummy model
… migrate to pyproject.toml and include stella/data/*.keras
…dels in shortest demo
…hrough in pipeline; build: add optional deps for jax/torch in pyproject
… NumPy 1.24+ ValueError
…o pyproject and remove setup.*
… dtype in medfilt/interp1d
…KEND, and stella.require_backend()
…rch); fix rotations Time/Quantity handling; modernize tqdm imports; remove hard-coded paths and Dropbox refs; docs pipeline uses packaged models - backends: require_backend() picks JAX if installed, else Torch; error only if neither installed; sets KERAS_BACKEND when auto-selecting - neural_network: call require_backend() before importing keras to avoid TF fallback - rotations: normalize time/flux/err to float arrays to avoid astropy Time/TimeDelta issues in LS - notebooks: switch to tqdm.auto; use stella.models for model paths; neutralize local paths - docs: pipeline.rst uses stella.models.get_model_path()/list_model_paths()
…to avoid large artifacts in commits
…d in .gitignore\n- Remove both from index with (files remain locally)
…Provide html_context['style'] default in conf.py to avoid Jinja UndefinedError\n- Switch to tqdm.auto in tutorial imports\n- Wrap as code to avoid RST target reference\n- Pass .value arrays to cnn.predict for robustness
…material theme, mkdocstrings, mkdocs-jupyter\n- Convert key pages to Markdown and include notebooks in nav\n- Update docs workflow to build via mkdocs and deploy site/\n- Add MkDocs deps to dev extras
…r on mkdocs.yml, ignore site/
…ux to avoid macOS MPS OOM
…remove paths filter)
Jax/PyTorch backend.
fix landing page
…+ positioned bars to avoid duplicates (2 bars total)
…bar layout (models + per-model)
…ly; use context managers; ensure final completion; prefer workspace import in notebook
Docs/notebooks in nav
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Hi Adina,
I hope you're well! As noted in Issue #32, recent versions of tensorflow don't play well with stella, and this has made it a bit of a challenge for some students to use.
As an experiment, I sat down with the GPT5 Copilot and tried agent mode for the first time... I hadn't tried it before, and it was actually a lot of work to use. Blame me for any bugs in there.
Anyway - as far as I can tell, stella now works with Jax and PyTorch backends and python 3.12. As far as I can tell on light curves I've already looked at, I get the same answers as before - but I don't have a working install of old stella on this machine, so I can't rigorously compare them. I have tested all the functions I have used but not the NN training, and the performance seems good. I've added a few convenience functions.
Let me know what you think!
All the best,
Ben