int-brain-lab / ibleatools / 29347155504
60%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 28
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 56.368% (+1.0%) from 55.333%
29347155504

push

github

web-flow 
Add 3D ACG compute + opt-in S3 loader support (#85)

* cells: add compute_3d_acgs (firing-rate x log-time 3D ACG)

Firing-rate-decile x log-time-lag 3D autocorrelogram, matching Han Yu's
NEMO/ICLR pipeline exactly: linear ACG via spikeinterface (same algorithm
as npyx.c4.fast_acg3d), resampled onto the log-time axis via the real
npyx.corr.convert_acg_log (not a port). Lives here rather than only in the
sdsc-slurms batch script so any dataset can be recomputed with the same
parameters for inference, not just the original brainwide map insertions.

npyx/spikeinterface are imported lazily inside the function so the rest of
this module stays usable without the [full] extra. Adds npyx to the [full]
extra; its scikit-learn<1.6.0 pin is overridden workspace-wide (root
pyproject.toml [tool.uv]) to stay within this package's own validated
1.6-1.8 scikit-learn range.

* cells: vendor npyx's log-ACG resampling instead of depending on npyx

npyx pulls in a fragile chain for the ~40 lines actually used
(convert_acg_log/smooth): a scikit-learn<1.6.0 pin conflicting with this
package's own validated 1.6-1.8 range, an unpinned ipython resolving to a
version that broke `import npyx` outright, and `import imp` (removed in
Python 3.12) in an eagerly-imported, unrelated submodule.

Move convert_acg_log/smooth_gaussian_axis0 into a new ephysatlas.iblnpyx
module (GPL-3.0, same as npyx — see its docstring) as a short-term vendor
while a consensus is worked out with the npyx maintainer. Drop npyx from
the [full] extra; compute_3d_acgs now only needs spikeinterface.

Add a small wiring test (dummy spike trains) for compute_3d_acgs.

* style: ruff format cells.py, iblnpyx.py, test_cells.py

* Add opt-in S3 download/read support for 3D ACG cell aggregates

Adds acg3d=True flag to download_cells_features/download_project_data
for clusters.acgs_3d.npy + acgs_3d.times.npy, and existence-gated
loading in read_cells_features. Mirrors the large_files pattern used
for waveform files.
... (continued)

1589 of 2819 relevant lines covered (56.37%)

0.56 hits per line

Coverage Regressions

Lines Coverage ∆ File
100
51.12
 0.74% data.py
71
53.21
 6.34% cells.py
Jobs
ID Job ID Ran Files Coverage
1 29347155504.1 28
56.37
 GitHub Action Run
Source Files on build 29347155504