Installation¶
Python 3.11 to 3.13. The package depends on geepers for UNR GNSS access and MIDAS rates, installed from GitHub.
With pip¶
pip install "disp2vlm-validation @ git+https://github.com/mgovorcin/disp2vlm_validation"
disp2vlm-validate --help
For development (pixi)¶
The repository has a pixi workspace with a locked environment:
git clone https://github.com/mgovorcin/disp2vlm_validation
cd disp2vlm_validation
pixi install -e dev
pixi run -e dev test # pytest
pixi run -e dev lint # pre-commit (ruff, black, mypy, SPDX headers)
pixi run -e dev docs # build this site (mkdocs, strict)
pixi run -e dev disp2vlm-validate gate
Optional: Venti¶
The venti extra installs Venti,
used only to reproduce a candidate calibration in the trade studies. The
validator itself does not need it.
Running full frames¶
A full DISP-S1 frame is about 7700 × 9500 pixels and 200–300 epochs. The
stack is streamed once per run, and the raw velocity is cached per stack, so a
run needs a few GB of memory, not the whole stack. On Linux hosts with
fragmented transparent huge pages, set NUMPY_MADVISE_HUGEPAGE=0: without it
large array allocations can be very slow.