gammalearn issueshttps://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues2024-02-21T15:54:56+01:00https://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/116wandb: upload settings2024-02-21T15:54:56+01:00Vuillaumewandb: upload settingsIt would be great to upload the `settings.py` file as artefact
And possibly a subset of settings as config.It would be great to upload the `settings.py` file as artefact
And possibly a subset of settings as config.https://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/115generate experiment settings script2023-12-18T17:16:07+01:00Vuillaumegenerate experiment settings scriptfollowing https://gitlab.in2p3.fr/gammalearn/gammalearn-data/-/tree/master/scripts/experiment_template
it would be nice to actually include that in gammalearn with an integrated command line to help users quickly setup an exp
~Improvementfollowing https://gitlab.in2p3.fr/gammalearn/gammalearn-data/-/tree/master/scripts/experiment_template
it would be nice to actually include that in gammalearn with an integrated command line to help users quickly setup an exp
~Improvementhttps://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/114build image from pytorch image2023-12-05T10:29:11+01:00Vuillaumebuild image from pytorch imagecould apparently lead to faster execution...could apparently lead to faster execution...https://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/113apply cleaning mask loaded from DL1 files2023-11-10T18:11:42+01:00Vuillaumeapply cleaning mask loaded from DL1 files```
from ctapipe.io import read_table
from lstchain.io.io import dl1_images_lstcam_key
tab = read_table('dl1_simtel_corsika_theta_23.161_az_260.739_run100.h5', path=dl1_images_lstcam_key)
tab['image_mask']
``````
from ctapipe.io import read_table
from lstchain.io.io import dl1_images_lstcam_key
tab = read_table('dl1_simtel_corsika_theta_23.161_az_260.739_run100.h5', path=dl1_images_lstcam_key)
tab['image_mask']
```Cleaning studyhttps://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/110Single inference implementation2023-03-28T18:18:59+02:00VuillaumeSingle inference implementationThere are currently two implementations in `gammalearn` and `gl_dl1_to_dl2` entry points, doubling maintenance.
Should be streamlined.There are currently two implementations in `gammalearn` and `gl_dl1_to_dl2` entry points, doubling maintenance.
Should be streamlined.VuillaumeVuillaume2023-04-17https://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/109simplify gitlab-ci docker building2023-01-12T15:26:12+01:00Vuillaumesimplify gitlab-ci docker building- [ ] docker build should be sequential
- [ ] simplify format by creating a dedicated `.gitlab-ci-dockers.yml``- [ ] docker build should be sequential
- [ ] simplify format by creating a dedicated `.gitlab-ci-dockers.yml``VuillaumeVuillaumehttps://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/108Add torchsummary package2022-12-07T11:38:36+01:00DELL'AIERA MichaëlAdd torchsummary packageAdd torchsummary to check models architectureAdd torchsummary to check models architectureDELL'AIERA MichaëlDELL'AIERA Michaëlhttps://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/107Add drop_last as a parameter of the experiment setting file2022-11-28T14:51:39+01:00DELL'AIERA MichaëlAdd drop_last as a parameter of the experiment setting fileDeepJDOT needs drop_last=True to ensure that the optimal transport plan has always the same size. But the other methods don't need this requirement.DeepJDOT needs drop_last=True to ensure that the optimal transport plan has always the same size. But the other methods don't need this requirement.https://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/106Include camera geometry in lightning checkpoints2023-01-26T11:31:57+01:00Mikael JacquemontInclude camera geometry in lightning checkpointsWould be very useful in production mode.
Possible hack: define geometry as a hyperparameter and use lightning save_hyperparameter method in moduleWould be very useful in production mode.
Possible hack: define geometry as a hyperparameter and use lightning save_hyperparameter method in modulehttps://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/105Create a pytorch WeightEnum for stable models [Production]2022-11-21T13:52:25+01:00Mikael JacquemontCreate a pytorch WeightEnum for stable models [Production]The trained weights of stable models could be handled with pytorch WeightEnum.The trained weights of stable models could be handled with pytorch WeightEnum.https://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/102Remove LightningModule from experiment runner2022-11-15T09:39:28+01:00Mikael JacquemontRemove LightningModule from experiment runnerThe LightningModule could be removed from experiment_runner.py for better readability of the code. As discussed with @vuillaumeThe LightningModule could be removed from experiment_runner.py for better readability of the code. As discussed with @vuillaumeDELL'AIERA MichaëlDELL'AIERA Michaëlhttps://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/99Remove label production from dataset and use dl1 parameters instead2022-11-09T15:55:05+01:00Mikael JacquemontRemove label production from dataset and use dl1 parameters instead@vuillaume@vuillaumeDELL'AIERA MichaëlDELL'AIERA Michaëlhttps://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/98ctapipe CameraGeometry "LSTCam" will be removed2022-11-09T15:45:34+01:00Mikael Jacquemontctapipe CameraGeometry "LSTCam" will be removedCurrently, we transform all the data (MC and LST1) into the geometry LSTCam for convenience:
- the pixels are organized by module
- this geometry is suitable to produce index matrix for indexedconv
However, this geometry will be removed...Currently, we transform all the data (MC and LST1) into the geometry LSTCam for convenience:
- the pixels are organized by module
- this geometry is suitable to produce index matrix for indexedconv
However, this geometry will be removed from future versions of ctapipe.VuillaumeVuillaumehttps://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/97Uncertainty balancing: set logvar coeff as dict2022-11-07T11:27:32+01:00Mikael JacquemontUncertainty balancing: set logvar coeff as dicthttps://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/93incompatibility with numpy 1.23 ?2022-07-26T11:06:08+02:00Vuillaumeincompatibility with numpy 1.23 ?numpy v1.23 comes with a fresh install and then leads to the following error when trying to import gammalearn:
```
AttributeError: module 'numpy' has no attribute 'asscalar'
```
Investigating...numpy v1.23 comes with a fresh install and then leads to the following error when trying to import gammalearn:
```
AttributeError: module 'numpy' has no attribute 'asscalar'
```
Investigating...VuillaumeVuillaumehttps://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/92CI env update: pytorch version2022-07-18T11:34:38+02:00Mikael JacquemontCI env update: pytorch versionIn the CI operations, when updating the base env mamba installs pytorch 1.11 (which causes a seg fault when using the profiler) instead of the latest version (1.12)
See https://gitlab.in2p3.fr/gammalearn/gammalearn/-/jobs/513179 for an ...In the CI operations, when updating the base env mamba installs pytorch 1.11 (which causes a seg fault when using the profiler) instead of the latest version (1.12)
See https://gitlab.in2p3.fr/gammalearn/gammalearn/-/jobs/513179 for an exampleVuillaumeVuillaumehttps://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/91try/except necessary?2022-10-07T10:52:39+02:00Vuillaumetry/except necessary?Comment from Max:
https://gitlab.in2p3.fr/gammalearn/gammalearn/-/blob/master/gammalearn/criterions.py#L45
> Writing:
> ```
> try:
> assert alt1.data.nelement() > 0
> except AssertError:
> log.exception()
> raise
> ```
>
> ...Comment from Max:
https://gitlab.in2p3.fr/gammalearn/gammalearn/-/blob/master/gammalearn/criterions.py#L45
> Writing:
> ```
> try:
> assert alt1.data.nelement() > 0
> except AssertError:
> log.exception()
> raise
> ```
>
> Is a very weird way of doing:
> ``
> if alt.data.nelement() == 0:
> raise ValueError()
> ```
>
> Do you do this because the assertion can be switched off by passing the optimization option for python?Mikael JacquemontMikael Jacquemonthttps://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/88Improve dataset file mode performance on MUST2022-04-07T11:50:50+02:00Mikael JacquemontImprove dataset file mode performance on MUSTOptions:
1. Measure performance after cluster rebuilding is finished
1. Try to merge dl1 files
1. Try to copy dl1 files on the worker node
1. Try to copy merged dl1 file on the worker node
1. Same as 3 and 4 but on the SSD scratch zoneOptions:
1. Measure performance after cluster rebuilding is finished
1. Try to merge dl1 files
1. Try to copy dl1 files on the worker node
1. Try to copy merged dl1 file on the worker node
1. Same as 3 and 4 but on the SSD scratch zoneMikael JacquemontMikael Jacquemonthttps://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/87Docker in Docker deactivated (gitlab)2022-03-21T22:30:01+01:00Mikael JacquemontDocker in Docker deactivated (gitlab)see: https://gitlab.in2p3.fr/cc-in2p3-gitlab/maintenance/-/issues/184see: https://gitlab.in2p3.fr/cc-in2p3-gitlab/maintenance/-/issues/184VuillaumeVuillaumehttps://gitlab.in2p3.fr/gammalearn/gammalearn/-/issues/83resume_from_checkpoint deprecated2022-11-22T16:33:18+01:00Vuillaumeresume_from_checkpoint deprecated```
LightningDeprecationWarning: Setting `Trainer(resume_from_checkpoint=)` is deprecated in v1.5 and will be removed in v1.7. Please pass `Trainer.fit(ckpt_path=)` directly instead.
``````
LightningDeprecationWarning: Setting `Trainer(resume_from_checkpoint=)` is deprecated in v1.5 and will be removed in v1.7. Please pass `Trainer.fit(ckpt_path=)` directly instead.
```VuillaumeVuillaume