Skip to content

Implement brightness-dependent scale for Normal and Student-t - #3

Open
ThaoHuynh94 wants to merge 14 commits into
mainfrom
mfx-shoebox-qi-posteriors
Open

ThaoHuynh94 wants to merge 14 commits into
mainfrom
mfx-shoebox-qi-posteriors

Conversation

@ThaoHuynh94

Copy link
Copy Markdown
Collaborator
  • Add the brightness-dependent g = 9.5 ADU/photon scale to the observation likelihood.
  • Use the same scale calculation for both Normal and Student-t likelihoods.
  • Configure Student-t with df = 4.0
  • Remove the previous Student-t global learned-scale mechanism.

Thao Huynh and others added 14 commits July 7, 2026 23:22
- New integrator.preprocess CLI and ChunkedDataModule for 100M+ reflection training
- refl_io.py: vectorized O(N log N) groupby, float32 clamping, deferred dials import
- pred_io.py: single parquet scan replaces 4 separate scans
- predict.py: auto-resolve metadata.npy from manifest.yaml for chunked configs
- profile_surrogates.py: fix _softplus_inverse overflow (math.expm1)
- MFX workflow: reorganized scripts/sbatch/configs/phil/, updated READMEs, 5 test files
- Add combinedmask_scale_merge_phil/, sbatch/, scripts/archive/ to repo

This branch has not been deployed

No deployments
Sign up for free to 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