VIPP 0.13.0a8¶
VIPP 0.13.0a8 removes the separate cuCIM installation path, makes every reviewed Windows GPU implementation available through the normal CuPy/CuPyX installation, and strengthens graph editing, GPU selection, image-axis interpretation, and segmentation cleanup.
This remains alpha software. Preserve original images and workflows, test representative inputs, and review important outputs before using them for scientific conclusions or publication.
Highlights¶
One normal Windows installer¶
The standard CPU and NVIDIA CUDA 13 choices remain in the same Windows installer. The CUDA choice now contains every current GPU provider; there is no cuCIM ZIP, local source build, private wheel, or second installer to run.
Rolling-Ball Background, Subtract Background, Measure Objects, and
Measure Objects + Intensity use reviewed CuPy implementations. Existing
workflows pinned to either former cuCIM basic-measurement implementation migrate
to the corresponding CuPy identity. A broad saved library:cucim preference
cannot be mapped safely to one implementation, so VIPP keeps it visibly
unavailable for review instead of guessing.
Better GPU decisions and evidence¶
- Find fastest pipeline… keeps completed timing and scientific-comparison evidence visible, including results where CPU and GPU are too close to name a safe winner.
- Applying measured assignments revalidates the graph, source revision, compute intent, actual assignment, and workload. A clean analysis is not rejected merely because the analysis worker updated its own evidence state.
- The compute window identifies the selected CUDA device and separates concise setup guidance from advanced diagnostic detail.
- Background processing and basic object measurements now use CuPy-only providers with CPU parity, cancellation, cleanup, and memory admission.
Workflow authoring and image handling¶
- Dropping a compatible node onto a highlighted wire inserts it between the connected nodes as one undoable graph edit.
- Remove Outliers cleans isolated bright or dark binary pixels using a Fiji-compatible neighborhood rule, with reviewed CPU and GPU variants.
- Intensity-and-contrast nodes show input and output histograms consistently.
- Integer inputs use integer-safe bounds where fractional values would be invalid, and slider ranges stay practical while typed fields retain a wider expert range.
- Image Source and Batch share the same explicit axis-declaration model. An
unknown leading axis such as
QYXcan be reviewed and interpreted asZYX. - Rescale Axes supports Z as well as Y and X. When metadata has no usable spatial axes, the node explains its inferred trailing Y/X interpretation.
Installation¶
Windows installer¶
Download the explicitly unsigned installer and checksum file from the official v0.13.0a8 GitHub release:
VIPP-Setup-0.13.0a8-Windows-x86_64-UNSIGNED.exeSHA256SUMS-Windows-0.13.0a8.txt
Verify the complete SHA-256 value before opening the installer. Windows reports Unknown publisher for this unsigned alpha and may require More info → Run anyway. Stop if the checksum differs or antivirus reports a threat; never disable Windows security.
The managed installer keeps CPU and CUDA installations side by side and owns only its fixed per-track folders, shortcuts, and uninstall records. The NVIDIA choice installs the pinned CuPy/CuPyX CUDA 13 stack directly—no optional provider installer follows it.
Manual installation¶
CPU on Windows, Linux, or macOS:
The reviewed native-Windows CUDA route uses CPython 3.12:
python -m pip install --upgrade "napari[pyqt6]>=0.6" "napari-vipp[gpu-cuda13]==0.13.0a8"
vipp-compute-doctor --track cuda13
Compatibility and migration¶
- VIPP accepts workflow schemas 3 and 4 and writes schema 4.
- Batch configurations and manifests remain schema 3.
- Cached scientific results are not stored in workflow JSON; recalculate and validate after upgrading.
- Generated Python remains pinned to the exact VIPP version that created it; regenerate exports after upgrading.
- Saved exact basic-measurement pins migrate from
cucim-measure-objects-basic-v1andcucim-measure-objects-intensity-basic-v1to theircupy-*replacements. - A saved broad
library:cucimpreference stays unavailable until the user chooses a current implementation explicitly. - Older release pages remain the authoritative installation record for older cuCIM-based environments. Do not reuse an old private wheel or provider ZIP in a 0.13.0a8 environment.
Scientific boundary¶
CPU remains the portable reference and fallback. GPU eligibility is operation- and workload-specific; VIPP does not silently cast an image or change a scientific parameter to make a node run on a GPU. The basic CuPy measurement providers cover the reviewed 2D/3D label and intensity region, while extended measurement groups and unsupported dtypes remain on CPU.
Machine-local timing can help choose an implementation but is not a portable performance claim. Preserve actual implementation IDs, fallback decisions, software versions, device information, parameters, and validation evidence for consequential analyses.
Release verification¶
The annotated tag, public prerelease, PyPI files, and all six GitHub assets resolve to the release commit and exact hashes below. Documentation validation and deployment are gated separately so this table does not make a self-referential claim about the page that contains it.
| Check | Exact release record |
|---|---|
Annotated v0.13.0a8 tag |
5a66ae9d1098ca5a8d409a4075c585692e3c3638 |
| Exact-main application CI | Passed: run 32584690313 |
| Windows installer build smoke | Passed: run 32585512509 |
| Public GitHub alpha prerelease | Passed: six exact assets; no cuCIM asset by design |
| Public PyPI wheel and source archive | Passed: workflow 32592042093; public digests match |
| Full RTX 5090 GPU qualification | Passed: cuda:0, 19 implementations, 24 evidence owners; aggregate SHA-256 0365366dc23750e000c6e9c4f8b384cdf706afdcb338ae3a9f80cfad3d1d8506 from source 7189cf40280d895b61b061f1468767164ccfbcf4 |
| Affected a7→a8 CUDA update | Passed: no retired-provider residue, clean pip check, Doctor 19/19, both CuPy measurement providers passed parity/lifecycle checks, and visible VIPP 0.13.0a8 launch |
Wheel napari_vipp-0.13.0a8-py3-none-any.whl |
77f18e4f34541847c93aa0a54b230a586b9967cc9a66d97370ea0f15ff593ec5 |
Source archive napari_vipp-0.13.0a8.tar.gz |
4e1a7b40669832ad8e19e7e0297bae0c3d19fc175f7a99ba3d5284425dbbb433 |
VIPP-Setup-0.13.0a8-Windows-x86_64-UNSIGNED.exe |
5b8233a05696efbf8fea7557012934385021ea3e9018befdd7789bf624740528 |
VIPP-Setup-0.13.0a8-Windows-x86_64-UNSIGNED-release.json |
cac0276a7b0e4556bf93ba1c7bb270f209460070906be061157f783beee33d58 |
VIPP-Setup-0.13.0a8-Windows-x86_64-UNSIGNED-THIRD-PARTY-NOTICES.txt |
95ca668d0977be347ab39af820f97652ed3565ed7ea4218205bb3a716a847e2d |
SHA256SUMS-Windows-0.13.0a8.txt |
93ad582d443dd95625b1d512d0f3b572d909676565eeb694d53d5b4ffe0b0ab6 |
The application release and PyPI project page are the canonical distribution locations. The absence of a separate cuCIM asset is intentional.