Published January 20, 2022
| Version v0.9
Software
Open
GammaLearn
Authors/Creators
- 1. Univ. Savoie Mont-Blanc, CNRS, LAPP
Description
GammaLearn is a collaborative project to apply deep learning to the analysis of low-level Imaging Atmospheric Cherenkov Telescopes such as CTA. It provides a framework to easily train and apply models from a configuration file. Learn more at https://purl.org/gammalearn
Notes
Release Notes:
- Allows label smoothing classification
- Modifies cleaning transform to enable cleaning mask as a channel in data
- Modifies the experiment setting examples to allow domain adaptation. This includes the creation of a data module train and test regrouping the corresponding dataset parameters, a new target for the domain classifier, a hand-designed LR Scheduler matching the pytorch API.
- Implements the DANN (Ganin et al., 2016) method for domain adaptation
- Changes pytorch lightning to version 1.6
- Upgrades network definition by taking backbone out of GammaPhysNet
- Add a merge option in the experiment setting file to merge dl2 files after training and testing
- Use both obs_id and event_id instead of solely event_id to select unique events while loading data
- Add a progress bar to file loading
Files
gammalearn.zip
Files
(3.6 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:834b2e1b0870217520bc7b6c586e9308
|
3.6 MB | Preview Download |
Additional details
Related works
- Is derived from
- Software: git+https://gitlab.in2p3.fr/gammalearn/gammalearn.git (URL)