Maximilian Linhoff
Maximilian Linhoff
That sounds like a `pip install --user matplotlib ` might shadow the matplotlib in the conda env. What is `python -c "import matplotlib; print(matplotlib.__file__)"`?
Looking closer, I see now that these are particular operations. `put` succeeded but `register` failed. Still, in this case I would have expected `"OK": False`.
Ok, if that's the general approach in DIRAC, that's fine I think then. Naively, I would expected a partially successful operation to have `Ok: False`, which would still then allow...
Why do std and median have different masks? Why is the median Nan unmasked but std masked?
It seems that this is not fully implemented yet. E..g. `Model.plot` seems to ignore this setting, reopening
Same for most of the IRF plotting methods, e.g. `EffectiveArea2D.plot_energy_depedence`
I allowed broadcasting now, but the tests still fail on shapes that we might think of as compatible but are not to the numpy broadcasting rules: ``` FAILED gammapy/maps/hpx/tests/test_ndmap.py::test_partial_hpx_map_cutout -...
Ah, btw.: I should mention that these are only the tests failing in `gammapy.maps`, running the whole test suite, many, many, many more tests are failing. I'll compile a list...
These are the unique shape combinations that make the tests fail: ``` (1000,), (1000, 1, 1) (100050,), (2001, 50, 1, 1) (100,), (100, 1, 1) (100, 3, 3), (100, 1,...