benchmarks

gammapy-benchmarks

Performance benchmarks for Gammapy.

Note that we have validation separately, see here.

Analyses

We have the following typical science cases tested:

• analysis_3d / script / results / 3d stacked analysis
• analysis_3d_joint / script / results - 3d joint analysis
• spectrum_1d / script / results - 1d stacked analysis
• spectrum_1d_joint / script / results - 1d joint analysis
• lightcurve_1d / script / results - 1d lightcurve analysis
• lightcurve_3d / script / results - 3d lightcurve analysis
• io / script / results - Read DL3 events data
• npred / script / results - Compute multiple npred calculations

Execution

To run a specific benchmark use make.py.

python make.py run-benchmark benchmark-name

To run all benchmarks:

python make.py run-benchmark all

To debug scripts (run quickly using only 2 observations):

GAMMAPY_BENCH_N_OBS=2 python make.py run-benchmark all

Results

A summary of the results (for 100 runs) can be found here.