-
Notifications
You must be signed in to change notification settings - Fork 77
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
cleanup and bugfix in viz dl2 #279
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Hi @vuillaut, thanks for the fixes.
Again, nothing against ctaplot
, but for me it is an overkill using external libraries to plot either roc curves, energy/angular resolutions or sensitivities. Since this is a personal package, I would like to hear the opinion from @kosack whether it is recommended or not to use it (why is it not used anywhere in ctapipe?)
I disagree there - the way in which sensitivity curves, IRFs, etc are computed and plotted, is quite complex and can be done in many ways with various assumptions that hide what the user is doing and make it impossible to do a true comparison. We've had many problems with that in the past - not everyone agrees even what "PSF" is (90% containment? RMS? etc). However, we've been discussing with @vuillaut and others that sensitivity (and other IRFs) must be in a common package, and will start to move toward that (and perhaps ctaplot is not the right place). The idea of ctaplot is to have a nice toolbox for making benchmark plots for code generated by ctapipe-based or other pipelines (that use a common data format and model), in a common way, with common assumptions, so that we can generate a set of notebooks with all the required benchmarks, without having a ton of boilerplate code pasted in each one. |
I am actually in the process of moving |
We will also use ctaplot in the Protopipe benchmarks |
I think these are two different things: shouldn't the "toolbox for making benchmark plots" be the purpose of
and I had already discussed about it at least in 3/4 PRs. We are calculating everything in
the definition of ctapipe is "CTA Low-level Data Processing Pipeline Framework Prototype" |
I thought the problem was adding it as a dependency but was fine to have it for control plots, which is exactly the case here.
And yet... 🙄 |
For the record, |
I think you're ignoring the word "framework" - its a library for making pipelines, not a pipeline |
Ok, sorry it took so long to get back to this. If we want to use |
any news on this @vuillaut? do you agree with the latest proposal? |
Sorry, I completely over-regarded this issue and then forgot.
I am not sure to understand what you mean here, apart from improving the docstring?
The computation and the plotting can be separated for some of the metrics (actually ctaplot is built that way and the functions exist). |
Connected to #391, ctaplot was already moved into the
I think there have been some documentation updates since this PR was opened, I'll have a look at that.
what I was proposing here is that this is actually written by Final question, are we compatible with the latest ctaplot release (v0.5.0)? We should probably follow some released version to avoid breaking of the code in the future. |
No. These file will be the input to ctaplot. Not its output. DL3 is photon lists AND the IRFs already included. See https://gamma-astro-data-formats.readthedocs.io/en/latest/ Reconstructed event lists used as input for the IRF calculation would be considered DL2. Work as also started on a common package for calculation of IRFs: https://github.com/cta-observatory/pyrf ctaplot should read input and plot, not make irf calculations or complicated IO. |
Sorry catching up with all the weekend exchanges. The definitive resolution should come at level 3 (DL3), once all cuts have been optimised, with the complete IRFs. pyrf is an attempt to do that (very preliminary) and the output should be DL3 FITS IRFs as defined in the gamma astro data format.
Yes, no problem to have the values and plots (maybe not for all).
That I think was the subject of this PR. Let me know, I can either update this PR or close it. |
btw, we agreed in #391 to add |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Just a couple of comments, also travis is complaining about lstchain_mc_rfperformance.py
, otherwise it looks good to go.
Codecov Report
@@ Coverage Diff @@
## master #279 +/- ##
==========================================
+ Coverage 43.95% 46.07% +2.12%
==========================================
Files 71 71
Lines 5192 5200 +8
==========================================
+ Hits 2282 2396 +114
+ Misses 2910 2804 -106
Continue to review full report at Codecov.
|
Fixes issue #278
I took this as an opportunity to do some cleanup in viz functions .