Dask hists #33
Dask hists #33
Conversation
|
Thanks @btovar! I see a few lint errors |
|
FYI I triggered a CI check in TopEFT using this branch. If that passes, then I'll be happy with merging this PR. |
|
@btovar looks like there's a |
|
There something weird with the tests, though. test_HistEFT_add should be failing... |
|
@btovar the CI/CD test that's being done looks to be |
Good catch! I just pushed that. Looks like the unit test is still for |
|
Looks like the CI is still failing. Seems to be a |
topcoffea/modules/histEFT.py
Outdated
|
|
||
| def _fill_indices(self, n_events): | ||
| # turns [0, 1, 2, ..., num of quadratic coeffs - 1] | ||
| # into: | ||
| # [0, 1, 2, ..., 0, 1, 2 ...,] | ||
| # repeated n_events times. | ||
| return self.array_backend.broadcast_to( | ||
| np.ogrid[0 : self.quad_count], (n_events, self.quad_count) | ||
| np.ogrid[0 : self.quad_count], (n_events, self.quad_count), |
There was a problem hiding this comment.
Is the comma needed? I guess it doesn't hurt anything.
|
@btovar do we need to add |
Needed because coffea is pinned to an older version
|
Ah yes, the old coffea does not add them. |
Ok, now it's an issue with the label keyword E TypeError: __init__() got an unexpected keyword argument 'label' |
|
@btovar looks like the current CI failure might just be a precision issue |
|
Thanks @btovar! Looks like the CI here is passing now. I've restarted the CI on TopEFT to make sure it works there as well. |
|
Nice! I need to update topEFT, though, and probably propagate changes to this pr. |
Ok thanks. It's complaining about |
|
I think so. For the new hists the axes need to be separated into lists of category or dense so that we know which axes are available at dask graph construction time. |
HistEFT to dask