You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
v4.x branch is the current generation (post-v3 architectural change to single-model cyto3)
Investigation needed
What cellpose version is fdf7a8c3a305a26e? The Wave container hash doesn't reveal the underlying pip-installed cellpose version. Need to either:
Decode by pulling the container and running cellpose --version
Look at the Wave build manifest, or
Trace the original Seqera Wave recipe
Are we on v3 or v4? v3 → v4 had architectural changes (single-model cyto3 instead of separate models). If we're on v3, this is a substantial test surface; if already v4, a minor bump.
Behavioral compatibility of our existing args (--flow_threshold 0, --batch_size 1) at the latest version.
Migration plan
Decode the Wave container to determine the current cellpose version
Review v4.1.1 changelog for behavioral changes affecting --flow_threshold / --batch_size / model selection
If a minor bump (within v4.x), update the Wave container; if v3 → v4, plan a model-switch and re-benchmark
Re-run image-mode tests on a Xenium v1 bundle to confirm segmentation quality matches
Risks
v3 → v4 architectural change is a notable test surface (different default models, different argument defaults)
GPU container variant: nf-core/cellpose may have separate CPU/GPU containers — bump consistently
Atera relevance: image mode is touched by Atera workflow but blocked by XR (out of scope per user direction); this upgrade primarily helps Xenium v1 image mode
Cross-links
Triggered by: Atera compatibility session 2026-05-28 (broader tool inventory check).
Current state
community.wave.seqera.io/library/python_pip_cellpose:fdf7a8c3a305a26eat modules/nf-core/cellpose/main.nf:9modules/nf-core/cellposeimage(when--method cellpose, the default)Latest upstream
cyto3)Investigation needed
fdf7a8c3a305a26e? The Wave container hash doesn't reveal the underlying pip-installed cellpose version. Need to either:cellpose --versioncyto3instead of separate models). If we're on v3, this is a substantial test surface; if already v4, a minor bump.--flow_threshold 0,--batch_size 1) at the latest version.Migration plan
--flow_threshold/--batch_size/ model selectionRisks
Cross-links