Skip to content

smoke_workspace_fixes.md

Latest commit

 

History

History
66 lines (50 loc) · 2.85 KB

smoke_workspace_fixes.md

File metadata and controls

66 lines (50 loc) · 2.85 KB
Raw

The new workspace smoke-test GitHub Actions (added via feature/smoke-test-ci) surfaced

Type: test Target: workspaces Difficulty: too-large Autonomy: supervised Priority: normal Status: formalised Filed: 2026-04-27 (backfilled from git)

The new workspace smoke-test GitHub Actions (added via feature/smoke-test-ci) surfaced pre-existing failures that were commented out to get the green baseline. Re-enable each entry once its underlying bug is fixed.

autofit_workspace/smoke_tests.txt

  • cookbooks/samples.py — IndexError: list index out of range in bypass mode (PYAUTO_TEST_MODE=2). Was already in the original smoke list but apparently never exercised in bypass mode before. Investigate which samples.* list access runs past its length when the sampler is skipped.

autogalaxy_workspace/smoke_tests.txt

  • group/modeling.py — committed dataset (150x150) fights with PYAUTO_SMALL_DATASETS=1 cap (15x15). Either add an unset: [PYAUTO_SMALL_DATASETS] override in config/build/env_vars.yaml for group/modeling, or delete the committed 150x150 dataset so the simulator regenerates it at 15x15.
  • multi/modeling.py — same pattern with a committed 100x100 dataset.
  • ellipse/fit.py — same pattern with the ellipse dataset.

Pick one approach (override vs strip-committed-data) and apply consistently.

autolens_workspace/smoke_tests.txt

  • imaging/likelihood_function.py — committed dataset (100x100) clashes with PYAUTO_SMALL_DATASETS=1. Same fix as the autogalaxy entries above.

autofit_workspace_test/smoke_tests.txt

  • searches/Emcee.py — TypeError: only 0-dimensional arrays can be converted to Python scalars on Python 3.13 + numpy 2.4. Failure is in autofit/non_linear/samples/mcmc.py:147 (median_pdf) when converting an ndarray result to a scalar. Needs a numpy-2.x-safe conversion (.item() on a single element, or explicit indexing).

autolens_workspace_test/smoke_tests.txt

  • jax_likelihood_functions/imaging/delaunay_mge.py — AttributeError: jax.interpreters.xla.pytype_aval_mappings was deprecated in JAX v0.5.0 and removed in JAX v0.7.0. The internal call site needs migrating to jax.core.pytype_aval_mappings. Likely somewhere in PyAutoArray or PyAutoGalaxy JAX-interop code.

Cross-cutting fixes already landed in this task

  • PyAutoFit PR #1219: made pandas.errors.SettingWithCopyWarning import tolerate pandas >= 2.2. This unblocks many autofit_workspace_test scripts (DynestyStatic, Emcee, minimal_output, latent) and autolens_workspace_test/database/scrape/general.py.
  • autolens_workspace_test CI installs tensorflow-probability==0.25.0 so Matérn-kernel JAX likelihood functions can import.
  • autofit_workspace_test CI installs nautilus-sampler so scripts/searches/Nautilus.py runs.