Dependency watch: only propose runner-installable Python versions; fail on silent pr-create errors - #4
Merged
Conversation
dev_maintenance.rst documents the six-step release routine the release workflow is built against. credits.rst carries the mbirtorch-2026 citation (Buzzard and Bouman); mbirjax-2024 and mbirtorch-2026 join refs.bib so both render under References; CITATION.cff gives GitHub its Cite-this-repository button. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…columns at a time, where each batch's full slice height is gathered from the owning GPUs and projected in one full-height kernel call
Replaces the single ported demo with the approved consolidated design (plans repo: plans/torch_port/active/demo_consolidation.md): parallel basics; cone beam with the Bouman-Sauer transmission noise model and weighting; the region-of-interest and axial field-of-view pairs, each shown without and with padding; direct vs MBIR at sparse views; helical; multiaxis (with generate_demo_data extended and tested); units and voxel shape; the denoiser. All at nominal 128 resolution, viewers opening at display minimum 0. The Demos and FAQs page carries the demo table and the rotation-direction FAQ paragraph. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
# Conflicts: # tests/test_demo_data.py
A bare direct_recon on an unconfigured model ran on one GPU: the automatic device selection lived only in vcd_recon, so MAR's first FDK put the whole volume on GPU 0 and segmentation OOMed (job 15047383). fdk_recon now calls _apply_device_policy() first -- a no-op when the user chose devices, the same call recon() makes otherwise. Note for Greg: as written this consults the widening floors calibrated on 3-iteration vcd (the A2 design question); change freely if a different rule fits FDK. The same latent gap exists in the other geometries' direct recons; left untouched pending this design call. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…it code Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…e it yet Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The watch trusted torch's wheel list alone, so it proposed 3.15, which setup-python cannot install yet (PR #3's failed CI job). The checker now reads setup-python's versions manifest and filters additions to stable releases the runners provide, reporting the rest informationally. An unreadable manifest is verdict UNKNOWN, not 'no divergence'. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…14-3.15 Add Python 3.13, 3.14 and 3.15 to the CI test matrix
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Two fixes from today's first live pull request (#3):
gh pr createfails the run instead of reporting success throughtee.Needs to be on main before tonight's scheduled run.
🤖 Generated with Claude Code