Skip to content

fix: stop clobbering working tfp-nightly in smoke installer (JAX Matern/delaunay_mge)#138

Merged
Jammy2211 merged 2 commits into
mainfrom
claude/pyautoconf-pyautonerves-rename-ik0cgu
Jul 19, 2026
Merged

fix: stop clobbering working tfp-nightly in smoke installer (JAX Matern/delaunay_mge)#138
Jammy2211 merged 2 commits into
mainfrom
claude/pyautoconf-pyautonerves-rename-ik0cgu

Conversation

@Jammy2211

Copy link
Copy Markdown
Collaborator

Problem

The JAX Matern-kernel (delaunay_mge) likelihood path fails at import with:

tensorflow_probability/.../jax/ops.py:681
    jax.interpreters.xla.pytype_aval_mappings[onp.ndarray])
AttributeError: module 'jax.interpreters.xla' has no attribute 'pytype_aval_mappings'

Root cause

.github/scripts/smoke_install.sh ran pip install tensorflow-probability==0.25.0 after PyAutoArray[optional], overwriting the working pinned tfp-nightly==0.26.0.dev20260713 (which provides the modified-Bessel bessel_kve) with the stable release. Stable tfp crashes at import under modern JAX (jax.interpreters.xla.pytype_aval_mappings was removed in JAX ≥ 0.10), as documented in PyAutoArray/pyproject.toml.

Fix

Remove the clobbering line and document why; the JAX path now imports the compatible tfp-nightly provided by PyAutoArray[optional].

🤖 Generated with Claude Code

https://claude.ai/code/session_013ciVftxvYpefh59wSkR7jN


Generated by Claude Code

claude added 2 commits July 19, 2026 15:05
The smoke installer ran `pip install tensorflow-probability==0.25.0` after
`PyAutoArray[optional]`, overwriting the working `tfp-nightly` (which
PyAutoArray[optional] pins) with the stable release. Stable tfp crashes at
import under the resolved modern JAX — it references
`jax.interpreters.xla.pytype_aval_mappings`, removed from JAX >=0.10 — which
broke the JAX Matern-kernel (`delaunay_mge`) likelihood path. Remove the line
and document why; the JAX path now imports the compatible tfp-nightly.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013ciVftxvYpefh59wSkR7jN
`jupyter nbconvert --execute` runs the kernel in the notebook's own directory
(e.g. `notebooks/interferometer/`), so a notebook that reads a committed dataset
via a repo-root-relative path (`Path("dataset", ...)`, matching the workspace
"run from the repo root" convention and the script smoke's `cwd=WORKSPACE`)
fails with FileNotFoundError. Stage a temporary copy of the notebook at the
workspace root before executing so the kernel's working directory is the root.

This fixes the `interferometer/simulator.ipynb` and `imaging/modeling.ipynb`
notebook smoke failures (missing `dataset/interferometer/uv_wavelengths/sma.fits`).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013ciVftxvYpefh59wSkR7jN
@Jammy2211
Jammy2211 merged commit 327edd2 into main Jul 19, 2026
8 checks passed
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.

2 participants