gregorkrzmanc commited on
Commit
e75a247
·
1 Parent(s): 6457dca
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .env +8 -0
  2. .gitignore +2 -0
  3. .idea/.gitignore +8 -0
  4. .idea/deployment.xml +46 -0
  5. .idea/inspectionProfiles/profiles_settings.xml +6 -0
  6. .idea/jetclustering.iml +14 -0
  7. .idea/jupyter-settings.xml +23 -0
  8. .idea/misc.xml +7 -0
  9. .idea/modules.xml +14 -0
  10. .idea/vcs.xml +12 -0
  11. Dockerfile +87 -0
  12. Dockerfile_training +77 -0
  13. README.md +65 -11
  14. app.py +68 -0
  15. config_files/config_jets.yaml +172 -0
  16. config_files/config_jets_1.yaml +191 -0
  17. config_files/config_jets_1_delphes.yaml +86 -0
  18. config_files/config_jets_2_delphes.yaml +63 -0
  19. container_shell.sh +4 -0
  20. docker-compose.yaml +7 -0
  21. env.sh +17 -0
  22. jobs/BigTraining_2_spatial_part_only_t3.slurm +15 -0
  23. jobs/BigTraining_2_spatial_part_only_vega.slurm +21 -0
  24. jobs/IRC_training/Delphes_training_t3_NoPID_augment.sh +24 -0
  25. jobs/IRC_training/Delphes_training_t3_NoPID_augment_IRC.sh +24 -0
  26. jobs/IRC_training/Delphes_training_t3_NoPID_augment_IRC_SN.sh +24 -0
  27. jobs/IRC_training/start_at_30k/Delphes_training_t3_NoPID_augment.sh +24 -0
  28. jobs/IRC_training/start_at_30k/Delphes_training_t3_NoPID_augment_IRC.sh +26 -0
  29. jobs/IRC_training/start_at_50k/Delphes_training_t3_NoPID_augment.sh +24 -0
  30. jobs/IRC_training/start_at_50k/Delphes_training_t3_NoPID_augment_IRC.sh +25 -0
  31. jobs/IRC_training/start_at_50k/Delphes_training_t3_NoPID_augment_IRCSN.sh +25 -0
  32. jobs/IRC_training/start_at_50k/Delphes_training_t3_NoPID_augment_IRC_noaug.sh +34 -0
  33. jobs/IRC_training/start_at_50k/test.sh +32 -0
  34. jobs/base_training/gatr_training_NoPIDDelphes.sh +26 -0
  35. jobs/base_training/lgatr_training_NoPIDDelphes.sh +26 -0
  36. jobs/base_training/transformer_training_NoPIDDelphes.sh +24 -0
  37. jobs/base_training_different_datasets/aug/lgatr_700_07.sh +26 -0
  38. jobs/base_training_different_datasets/aug/lgatr_700_07_and_900_03.sh +25 -0
  39. jobs/base_training_different_datasets/aug/lgatr_700_07_and_900_03_and_QCD.sh +25 -0
  40. jobs/base_training_different_datasets/aug/lgatr_QCD.sh +25 -0
  41. jobs/base_training_different_datasets/aug_IRC_S/lgatr_700_07.sh +27 -0
  42. jobs/base_training_different_datasets/aug_IRC_S/lgatr_700_07_and_900_03.sh +26 -0
  43. jobs/base_training_different_datasets/aug_IRC_S/lgatr_700_07_and_900_03_and_QCD.sh +25 -0
  44. jobs/base_training_different_datasets/aug_IRC_S/lgatr_QCD.sh +26 -0
  45. jobs/base_training_different_datasets/aug_IRC_SN/lgatr_700_07.sh +27 -0
  46. jobs/base_training_different_datasets/aug_IRC_SN/lgatr_700_07_and_900_03.sh +26 -0
  47. jobs/base_training_different_datasets/aug_IRC_SN/lgatr_700_07_and_900_03_and_QCD.sh +25 -0
  48. jobs/base_training_different_datasets/aug_IRC_SN/lgatr_900_03.sh +25 -0
  49. jobs/base_training_different_datasets/aug_IRC_SN/lgatr_QCD.sh +26 -0
  50. jobs/base_training_different_datasets/lgatr_700_07.sh +26 -0
.env ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ SVJ_CODE_ROOT=/work/gkrzmanc/jetclustering/code
2
+ SVJ_DATA_ROOT=/work/gkrzmanc/jetclustering/data
3
+ #SVJ_PREPROCESSED_DATA_ROOT=/pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc/jetclustering/preprocessed_data
4
+ SVJ_PREPROCESSED_DATA_ROOT=/work/gkrzmanc/jetclustering/preprocessed_data
5
+ SVJ_RESULTS_ROOT=/work/gkrzmanc/jetclustering/results
6
+ SVJ_RESULTS_ROOT_FALLBACK=/pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc/jetclustering/results
7
+ SVJ_WANDB_ENTITY=fcc_ml
8
+ WANDB_API_KEY=aaee0ebacafd9aac2eac525cae8d1b0919c6d9ec
.gitignore ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ models/
2
+ demo_datasets/
.idea/.gitignore ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ # Default ignored files
2
+ /shelf/
3
+ /workspace.xml
4
+ # Editor-based HTTP Client requests
5
+ /httpRequests/
6
+ # Datasource local storage ignored files
7
+ /dataSources/
8
+ /dataSources.local.xml
.idea/deployment.xml ADDED
@@ -0,0 +1,46 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0" encoding="UTF-8"?>
2
+ <project version="4">
3
+ <component name="PublishConfigData" autoUpload="Always" serverName="[email protected]:22 agent" remoteFilesAllowedToDisappearOnAutoupload="false" confirmBeforeUploading="false">
4
+ <option name="confirmBeforeUploading" value="false" />
5
+ <serverData>
6
+ <paths name="gkrzmanc@localhost:5555 password">
7
+ <serverdata>
8
+ <mappings>
9
+ <mapping deploy="/eos/home-g/gkrzmanc/jetclustering/code" local="$PROJECT_DIR$" web="" />
10
+ </mappings>
11
+ </serverdata>
12
+ </paths>
13
+ <paths name="gkrzmanc@localhost:5555 password (2)">
14
+ <serverdata>
15
+ <mappings>
16
+ <mapping deploy="/eos/home-g/gkrzmanc/jetclustering/code" local="$PROJECT_DIR$" web="/" />
17
+ </mappings>
18
+ </serverdata>
19
+ </paths>
20
+ <paths name="[email protected]:22 password">
21
+ <serverdata>
22
+ <mappings>
23
+ <mapping deploy="/work/gkrzmanc/jetclustering/code" local="$PROJECT_DIR$" />
24
+ </mappings>
25
+ </serverdata>
26
+ </paths>
27
+ <paths name="[email protected]:22 password (2)">
28
+ <serverdata>
29
+ <mappings>
30
+ <mapping deploy="/work/gkrzmanc/jetclustering/code" local="$PROJECT_DIR$" />
31
+ <mapping deploy="/work/gkrzmanc/CMSSW_10_6_26/src/PhysicsTools/SVJScouting" local="$PROJECT_DIR$/../SVJScouting_ntuplizer" />
32
+ <mapping deploy="/work/gkrzmanc/jetclustering/LJP" local="$PROJECT_DIR$/../LJP" />
33
+ </mappings>
34
+ </serverdata>
35
+ </paths>
36
+ <paths name="[email protected]:22 agent">
37
+ <serverdata>
38
+ <mappings>
39
+ <mapping deploy="/work/gkrzmanc/jetclustering/code" local="$PROJECT_DIR$" />
40
+ </mappings>
41
+ </serverdata>
42
+ </paths>
43
+ </serverData>
44
+ <option name="myAutoUpload" value="ALWAYS" />
45
+ </component>
46
+ </project>
.idea/inspectionProfiles/profiles_settings.xml ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ <component name="InspectionProjectProfileManager">
2
+ <settings>
3
+ <option name="USE_PROJECT_PROFILE" value="false" />
4
+ <version value="1.0" />
5
+ </settings>
6
+ </component>
.idea/jetclustering.iml ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0" encoding="UTF-8"?>
2
+ <module type="PYTHON_MODULE" version="4">
3
+ <component name="NewModuleRootManager">
4
+ <content url="file://$MODULE_DIR$" />
5
+ <orderEntry type="jdk" jdkName="Remote Python 3.10.16 (sftp://[email protected]:22/work/gkrzmanc/1gatr/bin/python)" jdkType="Python SDK" />
6
+ <orderEntry type="sourceFolder" forTests="false" />
7
+ <orderEntry type="module" module-name="lorentz-gatr" />
8
+ <orderEntry type="module" module-name="SVJScouting_ntuplizer" />
9
+ <orderEntry type="module" module-name="SVJProduction" />
10
+ <orderEntry type="module" module-name="lpc-scripts" />
11
+ <orderEntry type="module" module-name="LJP" />
12
+ <orderEntry type="module" module-name="delphes-python" />
13
+ </component>
14
+ </module>
.idea/jupyter-settings.xml ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0" encoding="UTF-8"?>
2
+ <project version="4">
3
+ <component name="JupyterPersistentConnectionParameters">
4
+ <option name="moduleParameters">
5
+ <map>
6
+ <entry key="$PROJECT_DIR$/.idea/jetclustering.iml">
7
+ <value>
8
+ <JupyterConnectionParameters>
9
+ <option name="managed" value="true" />
10
+ <option name="sdkName" value="Remote Python 3.10.16 (sftp://[email protected]:22/work/gkrzmanc/1gatr/bin/python)" />
11
+ <option name="environmentVariables">
12
+ <map>
13
+ <entry name="LD_LIBRARY_PATH" value="/eos/home-g/gkrzmanc/miniforge3/lib:/home/gkrzmanc/env/lib:" />
14
+ <entry name="PYTHONUNBUFFERED" value="1" />
15
+ </map>
16
+ </option>
17
+ </JupyterConnectionParameters>
18
+ </value>
19
+ </entry>
20
+ </map>
21
+ </option>
22
+ </component>
23
+ </project>
.idea/misc.xml ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0" encoding="UTF-8"?>
2
+ <project version="4">
3
+ <component name="Black">
4
+ <option name="sdkName" value="Python 3.10" />
5
+ </component>
6
+ <component name="ProjectRootManager" version="2" project-jdk-name="Remote Python 3.10.16 (sftp://[email protected]:22/work/gkrzmanc/1gatr/bin/python)" project-jdk-type="Python SDK" />
7
+ </project>
.idea/modules.xml ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0" encoding="UTF-8"?>
2
+ <project version="4">
3
+ <component name="ProjectModuleManager">
4
+ <modules>
5
+ <module fileurl="file://$PROJECT_DIR$/../LJP/.idea/LJP.iml" filepath="$PROJECT_DIR$/../LJP/.idea/LJP.iml" />
6
+ <module fileurl="file://$PROJECT_DIR$/../SVJProduction/.idea/SVJProduction.iml" filepath="$PROJECT_DIR$/../SVJProduction/.idea/SVJProduction.iml" />
7
+ <module fileurl="file://$PROJECT_DIR$/../SVJScouting_ntuplizer/.idea/SVJScouting_ntuplizer.iml" filepath="$PROJECT_DIR$/../SVJScouting_ntuplizer/.idea/SVJScouting_ntuplizer.iml" />
8
+ <module fileurl="file://$PROJECT_DIR$/../delphes-python/.idea/delphes-python.iml" filepath="$PROJECT_DIR$/../delphes-python/.idea/delphes-python.iml" />
9
+ <module fileurl="file://$PROJECT_DIR$/.idea/jetclustering.iml" filepath="$PROJECT_DIR$/.idea/jetclustering.iml" />
10
+ <module fileurl="file://$PROJECT_DIR$/../lorentz-gatr/.idea/lorentz-gatr.iml" filepath="$PROJECT_DIR$/../lorentz-gatr/.idea/lorentz-gatr.iml" />
11
+ <module fileurl="file://$PROJECT_DIR$/../lpc-scripts/.idea/lpc-scripts.iml" filepath="$PROJECT_DIR$/../lpc-scripts/.idea/lpc-scripts.iml" />
12
+ </modules>
13
+ </component>
14
+ </project>
.idea/vcs.xml ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0" encoding="UTF-8"?>
2
+ <project version="4">
3
+ <component name="VcsDirectoryMappings">
4
+ <mapping directory="" vcs="Git" />
5
+ <mapping directory="$PROJECT_DIR$/../LJP" vcs="Git" />
6
+ <mapping directory="$PROJECT_DIR$/../SVJProduction" vcs="Git" />
7
+ <mapping directory="$PROJECT_DIR$/../SVJScouting_ntuplizer" vcs="Git" />
8
+ <mapping directory="$PROJECT_DIR$/../delphes-python" vcs="Git" />
9
+ <mapping directory="$PROJECT_DIR$/../lorentz-gatr" vcs="Git" />
10
+ <mapping directory="$PROJECT_DIR$/../lpc-scripts" vcs="Git" />
11
+ </component>
12
+ </project>
Dockerfile ADDED
@@ -0,0 +1,87 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # gkrz/lgatr:v3
2
+ # docker build -t gkrz/lgatr:v4 .
3
+ FROM nvidia/cuda:11.8.0-runtime-ubuntu22.04
4
+
5
+ WORKDIR /app
6
+
7
+ COPY . /app
8
+
9
+ SHELL ["/bin/bash", "-c"]
10
+
11
+ USER root
12
+
13
+ RUN apt update && \
14
+ DEBIAN_FRONTEND=noninteractive apt install --yes --no-install-recommends \
15
+ build-essential \
16
+ cmake \
17
+ ffmpeg \
18
+ git \
19
+ python-is-python3 \
20
+ python3-dev \
21
+ python3-pip \
22
+ && \
23
+ rm -rf /var/lib/apt/lists/*
24
+
25
+ RUN python3.10 --version
26
+ RUN python3 --version
27
+ RUN python --version
28
+
29
+ RUN python3 -m pip install --no-cache-dir --upgrade pip
30
+ #python3 -m pip install --no-cache-dir --upgrade --requirement requirements.txt
31
+ RUN python3 -m pip install numba==0.58.1
32
+ # packages without conda
33
+ # RUN python3 -m pip install --no-cache-dir torch==1.13.1+cu117 torchvision==0.14.1+cu117 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu117
34
+ RUN python3 -m pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
35
+ RUN python3 -m pip install torch_geometric
36
+ RUN python3 -m pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.5.1+cu118.html
37
+ RUN python3 -m pip install pytorch-lightning yacs torchmetrics
38
+ RUN python3 -m pip install performer-pytorch
39
+ RUN python3 -m pip install tensorboardX
40
+ RUN python3 -m pip install ogb
41
+ RUN python3 -m pip install wandb
42
+ RUN python3 -m pip install seaborn
43
+ RUN python3 -m pip install dgl -f https://data.dgl.ai/wheels/cu118/repo.html
44
+ RUN python3 -m pip install numpy
45
+ RUN python3 -m pip install scipy
46
+ RUN python3 -m pip install pandas
47
+ RUN python3 -m pip install scikit-learn
48
+ RUN python3 -m pip install matplotlib
49
+ RUN python3 -m pip install tqdm
50
+ RUN python3 -m pip install PyYAML
51
+ RUN python3 -m pip install awkward0
52
+ RUN python3 -m pip install uproot
53
+ RUN python3 -m pip install awkward
54
+ RUN python3 -m pip install vector
55
+ RUN python3 -m pip install lz4
56
+ RUN python3 -m pip install xxhash
57
+ RUN python3 -m pip install tables
58
+ RUN python3 -m pip install tensorboard
59
+ RUN python3 -m pip install plotly
60
+ RUN python3 -m pip install xformers --index-url https://download.pytorch.org/whl/cu118
61
+ RUN python3 -m pip install fastjet
62
+ RUN python3 -m pip install gradio
63
+ RUN python3 -m pip install huggingface_hub
64
+
65
+ # remove pip cache
66
+ RUN python3 -m pip cache purge
67
+
68
+ # COPY docker/ext_packages /docker/ext_packages
69
+ # RUN python3 /docker/ext_packages/install_upstream_python_packages.py
70
+ RUN mkdir -p /opt/pepr
71
+
72
+ # Install GATr
73
+ #RUN cd /opt/pepr && git clone https://github.com/Qualcomm-AI-research/geometric-algebra-transformer.git geometric-algebra-transformer1
74
+ #RUN cd /opt/pepr/geometric-algebra-transformer1/ && python3 -m pip install .
75
+
76
+ # Install L-GATr - for some reason this only works if executed from the already-built container
77
+ RUN cd /opt/pepr && git clone https://github.com/gregorkrz/lorentz-gatr lgatr
78
+ RUN cd /opt/pepr/lgatr/ && python3 -m pip install .
79
+
80
+ # Install torch_cmspepr
81
+
82
+ RUN cd /opt/pepr && git clone https://github.com/cms-pepr/pytorch_cmspepr
83
+ RUN cd /opt/pepr/pytorch_cmspepr/ && python3 -m pip install .
84
+ RUN cd /root
85
+ RUN ls
86
+ # entrypoint run app.py with python
87
+ ENTRYPOINT ["python", "app.py"]
Dockerfile_training ADDED
@@ -0,0 +1,77 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # gkrz/lgatr:v3
2
+ # docker build -t gkrz/lgatr:v4 .
3
+ FROM nvidia/cuda:11.8.0-runtime-ubuntu22.04
4
+
5
+ SHELL ["/bin/bash", "-c"]
6
+
7
+ USER root
8
+
9
+ RUN apt update && \
10
+ DEBIAN_FRONTEND=noninteractive apt install --yes --no-install-recommends \
11
+ build-essential \
12
+ cmake \
13
+ ffmpeg \
14
+ git \
15
+ python-is-python3 \
16
+ python3-dev \
17
+ python3-pip \
18
+ && \
19
+ rm -rf /var/lib/apt/lists/*
20
+
21
+ RUN python3.10 --version
22
+ RUN python3 --version
23
+ RUN python --version
24
+
25
+ RUN python3 -m pip install --no-cache-dir --upgrade pip
26
+ #python3 -m pip install --no-cache-dir --upgrade --requirement requirements.txt
27
+ RUN python3 -m pip install numba==0.58.1
28
+ # packages without conda
29
+ # RUN python3 -m pip install --no-cache-dir torch==1.13.1+cu117 torchvision==0.14.1+cu117 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu117
30
+ RUN python3 -m pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
31
+ RUN python3 -m pip install torch_geometric
32
+ RUN python3 -m pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.5.1+cu118.html
33
+ RUN python3 -m pip install pytorch-lightning yacs torchmetrics
34
+ RUN python3 -m pip install performer-pytorch
35
+ RUN python3 -m pip install tensorboardX
36
+ RUN python3 -m pip install ogb
37
+ RUN python3 -m pip install wandb
38
+ RUN python3 -m pip install seaborn
39
+ RUN python3 -m pip install dgl -f https://data.dgl.ai/wheels/cu118/repo.html
40
+ RUN python3 -m pip install numpy
41
+ RUN python3 -m pip install scipy
42
+ RUN python3 -m pip install pandas
43
+ RUN python3 -m pip install scikit-learn
44
+ RUN python3 -m pip install matplotlib
45
+ RUN python3 -m pip install tqdm
46
+ RUN python3 -m pip install PyYAML
47
+ RUN python3 -m pip install awkward0
48
+ RUN python3 -m pip install uproot
49
+ RUN python3 -m pip install awkward
50
+ RUN python3 -m pip install vector
51
+ RUN python3 -m pip install lz4
52
+ RUN python3 -m pip install xxhash
53
+ RUN python3 -m pip install tables
54
+ RUN python3 -m pip install tensorboard
55
+ RUN python3 -m pip install plotly
56
+ RUN python3 -m pip install xformers --index-url https://download.pytorch.org/whl/cu118
57
+ RUN python3 -m pip install fastjet
58
+
59
+ # remove pip cache
60
+ RUN python3 -m pip cache purge
61
+
62
+ # COPY docker/ext_packages /docker/ext_packages
63
+ # RUN python3 /docker/ext_packages/install_upstream_python_packages.py
64
+ RUN mkdir -p /opt/pepr
65
+
66
+ # Install GATr
67
+ RUN cd /opt/pepr && git clone https://github.com/Qualcomm-AI-research/geometric-algebra-transformer.git geometric-algebra-transformer1
68
+ RUN cd /opt/pepr/geometric-algebra-transformer1/ && python3 -m pip install .
69
+
70
+ # Install L-GATr - for some reason this only works if executed from the already-built container
71
+ #RUN cd /opt/pepr && git clone https://github.com/gregorkrz/lorentz-gatr lgatr
72
+ #RUN cd /opt/pepr/lgatr/ && python3 -m pip install .
73
+
74
+ # Install torch_cmspepr
75
+
76
+ RUN cd /opt/pepr && git clone https://github.com/cms-pepr/pytorch_cmspepr
77
+ RUN cd /opt/pepr/pytorch_cmspepr/ && python3 -m pip install .
README.md CHANGED
@@ -1,11 +1,65 @@
1
- ---
2
- title: Jetclustering
3
- emoji: 🐢
4
- colorFrom: pink
5
- colorTo: gray
6
- sdk: docker
7
- pinned: false
8
- short_description: L-GATr based jet clustering
9
- ---
10
-
11
- Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # SVJ clustering
2
+ The repo has evolved from [here](https://github.com/selvaggi/mlpf) - mainly, we use the dataloader and code for reading the root files for the previous MLPF project. The preprocessing part is not really needed but it does help with performance when we are doing a lot of experiments with the same dataset.
3
+
4
+ ## Setup
5
+ **Important**: To make it easier and less time-consuming to move the commands across different machines, i.e. lxplus, T3 and Vega, we use relative paths. However, all commands can also be supplied absolute paths starting with `/`. **In case you use relative paths, make sure to modify the `env.sh` file with your paths!**
6
+
7
+ 0. Environment setup: We use the Python with packages compiled in the following container: `gkrz/lgatr:v3`. The container can be built from scratch using the Dockerfile in this repo.
8
+
9
+
10
+ 1. Set the environment variables `source env.sh`
11
+
12
+
13
+ ### Preprocess data
14
+ See the script at `sbatch jobs/preprocess_v0.slurm` (make sure to update your local `env.sh` file!)
15
+
16
+ ## Evaluation of clustering
17
+
18
+ For AK8: `python -m scripts.analysis.count_matched_quarks --input scouting_PFNano_signals/SVJ_hadronic_std --dataset-cap 1000`
19
+
20
+
21
+ For AK8 GenJets: `python -m scripts.analysis.count_matched_quarks --input scouting_PFNano_signals/SVJ_hadronic_std --dataset-cap 1000 --jets-object genjets`
22
+
23
+
24
+ For any model: `python -m scripts.analysis.count_matched_quarks --input scouting_PFNano_signals/SVJ_hadronic_std --output scouting_PFNano_signals2/SVJ_hadronic_std/all_models_eval/GATr_rinv_03_m_900 --eval-dir train/Test_betaPt_BC_all_datasets_2025_01_07_17_50_45 --dataset-cap 1000 --jets-object model_jets` Add `--eval-dir` with the path to the eval run containing the coordinates and clustering labels. Optionally, add `--clustering-suffix` in case there are multiple clusterings saved in the same folder. (usually not unless you were fine-tuning the clustering)
25
+
26
+
27
+ The script produces output in the `results` folder. The script goes over the events up to dataset-cap (optional).
28
+
29
+
30
+
31
+ ### Automated evaluation
32
+ In order to move things faster, scripts to evaluate the trained models faster at a given ckpt are given.
33
+
34
+ To evaluate at step 10k of the given training run: `python -m scripts.generate_test_jobs -template t3 -run Transformer_training_40k_5_64_4_2025_01_22_15_55_39 -step 10000 -tag params_study`
35
+ * Important: The step provided counts from the starting point of training the model: for example, if the run breaks in the middle and it's restarted from the latest ckpt, the command will identify that and load a checkpoint from the previous run if it contains one. You only need to provide the latest training with the `-run` argument.
36
+ * The `-tag` argument identifies the given study and can be later used to retrieve all the evals of all the models for a given run.
37
+ * The command pulls the config (e.g. model architecture and hyperparameters) automatically from the wandb run of the training.
38
+ * Add `-os` argument with a path to the objectness score checkpoint to use in the evaluation.
39
+
40
+
41
+ After the GPU eval, the CPU eval from above needs to be ran: `python -m scripts.test_plot_jobs --tag params_study`. The script will identify the runs that need to have evaluation figures produced. Uncommend the AK8 part in the file to also evaluate with AK8. Inside the produced folder, it also produces run_config.pkl that can be used later to make plots (of e.g. metrics vs number of params, model architecture, and amount of training).
42
+
43
+
44
+
45
+
46
+ Use the scripts in `scripts/` to produce the joint plots of F1 score, precision, recall etc.
47
+
48
+
49
+
50
+
51
+ ## Training
52
+
53
+ See mainly `jobs/vega/lgatr_training.sh`, `jobs/vega/transformer_training.sh`, `jobs/vega/gatr_training_vega.sh` - you might need to modify the slurm file a bit to fit the system you are running on
54
+
55
+
56
+
57
+ ### Datasets
58
+
59
+ `scouting_PFNano_signals1`: Contains special PFCands and PFCands in separate fields
60
+
61
+ `scouting_PFNano_signals2`: Contains both special PFCands and PFCands in the same field, under PFCands.
62
+
63
+ It was easier to just create this instead of always having special treatment for the special PFCands. As of January 2025, we are only using this version, accessible at `/pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc/jetclustering/preprocessed_data/scouting_PFNano_signals2`.
64
+
65
+
app.py ADDED
@@ -0,0 +1,68 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import gradio as gr
2
+ import matplotlib.pyplot as plt
3
+ import numpy as np
4
+ import os
5
+
6
+ from src.model_wrapper_gradio import inference
7
+
8
+ # === Dummy file-based prefill function ===
9
+ def prefill_event(subdataset, event_idx):
10
+ base_path = f"demo_datasets/{subdataset}/{event_idx}"
11
+ try:
12
+ with open(f"{base_path}.txt", "r") as f:
13
+ particles_data = f.read()
14
+ except FileNotFoundError:
15
+ particles_data = "pt eta phi mass charge\n"
16
+
17
+ try:
18
+ with open(f"{base_path}_quarks.txt", "r") as f:
19
+ quarks_data = f.read()
20
+ except FileNotFoundError:
21
+ quarks_data = "pt eta phi\n"
22
+
23
+ return particles_data, quarks_data
24
+
25
+
26
+ from huggingface_hub import snapshot_download
27
+
28
+ snapshot_download(repo_id="gregorkrzmanc/jetclustering", local_dir="src/models/")
29
+ snapshot_download(repo_id="gregorkrzmanc/jetclustering_demo", local_dir="demo_datasets/", repo_type="dataset")
30
+
31
+ # === Interface layout ===
32
+ def gradio_ui():
33
+ with gr.Blocks() as demo:
34
+ gr.Markdown("## Jet Clustering Demo")
35
+
36
+ with gr.Row():
37
+ loss_dropdown = gr.Dropdown(choices=["GP_IRC_SN", "GP_IRC_S", "GP", "base"], label="Loss Function", value="GP_IRC_SN")
38
+ train_dataset_dropdown = gr.Dropdown(choices=["QCD", "900_03", "900_03+700_07", "700_07", "900_03+700_07+QCD"], label="Training Dataset", value="QCD")
39
+
40
+ with gr.Row():
41
+ subdataset_dropdown = gr.Dropdown(choices=os.listdir("demo_datasets"), label="Subdataset")
42
+ event_idx_dropdown = gr.Dropdown(choices=list(range(50)), label="Event Index")
43
+ prefill_btn = gr.Button("Load Event from Dataset")
44
+
45
+ particles_text = gr.Textbox(label="Particles CSV (pt eta phi mass charge)", lines=6, interactive=True)
46
+ quarks_text = gr.Textbox(label="Quarks CSV (pt eta phi)", lines=3, interactive=True)
47
+
48
+ process_btn = gr.Button("Run Jet Clustering")
49
+
50
+ image_output = gr.Plot(label="Output")
51
+ model_jets_output = gr.JSON(label="Model Jets")
52
+ antikt_jets_output = gr.JSON(label="Anti-kt Jets")
53
+
54
+ prefill_btn.click(fn=prefill_event,
55
+ inputs=[subdataset_dropdown, event_idx_dropdown],
56
+ outputs=[particles_text, quarks_text])
57
+
58
+
59
+ process_btn.click(fn=inference,
60
+ inputs=[loss_dropdown, train_dataset_dropdown, particles_text, quarks_text],
61
+ outputs=[model_jets_output, antikt_jets_output, image_output])
62
+
63
+ return demo
64
+
65
+
66
+ demo = gradio_ui()
67
+ demo.launch()
68
+
config_files/config_jets.yaml ADDED
@@ -0,0 +1,172 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ treename: mmtree/Events;1
2
+ selection:
3
+ ### use `&`, `|`, `~` for logical operations on numpy arrays
4
+ ### can use functions from `math`, `np` (numpy), and `awkward` in the expression
5
+ #(jet_tightId==1) & (jet_no<2) & (fj_pt>200) & (fj_pt<2500) & (((sample_isQCD==0) & (fj_isQCD==0)) | ((sample_isQCD==1) & (fj_isQCD==1))) & (event_no%7!=0)
6
+ #(recojet_e>=5)
7
+
8
+ test_time_selection:
9
+ ### selection to apply at test time (i.e., when running w/ --predict)
10
+ #(jet_tightId==1) & (jet_no<2) & (fj_pt>200) & (fj_pt<2500) & (((sample_isQCD==0) & (fj_isQCD==0)) | ((sample_isQCD==1) & (fj_isQCD==1))) & (event_no%7==0)
11
+ #(recojet_e<5)
12
+
13
+ new_variables:
14
+ ### [format] name: formula
15
+ ### can use functions from `math`, `np` (numpy), and `awkward` in the expression
16
+ #pfcand_mask: awkward.JaggedArray.ones_like(pfcand_etarel)
17
+ #sv_mask: awkward.JaggedArray.ones_like(sv_etarel)
18
+ #pfcand_mask: awkward.JaggedArray.ones_like(pfcand_e)
19
+
20
+ preprocess:
21
+ ### method: [manual, auto] - whether to use manually specified parameters for variable standardization
22
+ ### [note]: `[var]_mask` will not be transformed even if `method=auto`
23
+
24
+ inputs:
25
+ n_fat_jets:
26
+ pad_mode: wrap
27
+ length: 1
28
+ vars:
29
+ - [nFatJet, null]
30
+ # - [nJetId, null]
31
+ fat_jets:
32
+ pad_mode: wrap
33
+ length: 50
34
+ vars:
35
+ - [FatJet_pt, null]
36
+ - [FatJet_eta, null]
37
+ - [FatJet_phi, null]
38
+ - [FatJet_mass, null]
39
+ n_jets:
40
+ pad_mode: wrap
41
+ length: 1
42
+ vars:
43
+ - [ nJet, null ]
44
+ jets:
45
+ pad_mode: wrap
46
+ length: 50
47
+ vars:
48
+ - [ Jet_pt, null ]
49
+ - [ Jet_eta, null ]
50
+ - [ Jet_phi, null ]
51
+ - [ Jet_mass, null ]
52
+ n_genjets:
53
+ pad_mode: wrap
54
+ length: 1
55
+ vars:
56
+ - [n_genjet, null]
57
+ genjets:
58
+ pad_mode: wrap
59
+ length: 50
60
+ vars:
61
+ - [GenFatJet_pt, null]
62
+ - [GenFatJet_eta, null]
63
+ - [GenFatJet_phi, null]
64
+ - [GenFatJet_mass, null]
65
+ n_pfcands:
66
+ pad_mode: wrap
67
+ length: 1
68
+ vars:
69
+ - [ nPFCands, null ]
70
+ pfcands:
71
+ pad_mode: wrap
72
+ length: 750
73
+ vars:
74
+ - [PFCands_pt, null]
75
+ - [PFCands_eta, null]
76
+ - [PFCands_phi, null]
77
+ - [PFCands_mass, null]
78
+ - [PFCands_charge, null]
79
+ - [PFCands_pdgId, null]
80
+
81
+ pfcands_jet_mapping:
82
+ pad_mode: wrap
83
+ length: 750
84
+ vars:
85
+ - [ FatJetPFCands_jetIdx, null ]
86
+ - [ FatJetPFCands_pFCandsIdx, null ]
87
+ #n_offline_pfcands:
88
+ # pad_mode: wrap
89
+ # length: 1
90
+ # vars:
91
+ # - [ nOfflinePFCands, null ]
92
+ #offline_pfcands:
93
+ # pad_mode: wrap
94
+ # length: 750
95
+ # vars:
96
+ # - [ OfflinePFCands_pt, null ]
97
+ # - [ OfflinePFCands_eta, null ]
98
+ # - [ OfflinePFCands_phi, null ]
99
+ # - [ OfflinePFCands_mass, null ]
100
+ # - [ OfflinePFCands_charge, null ]
101
+ # - [ OfflinePFCands_pdgId, null ]
102
+ #offline_pfcands_jet_mapping:
103
+ # pad_mode: wrap
104
+ # length: 750
105
+ # vars:
106
+ # - [ OfflineFatJetPFCands_jetIdx, null ]
107
+ # - [ OfflineFatJetPFCands_pFCandsIdx, null ]
108
+ MET:
109
+ pad_mode: wrap
110
+ length: 1
111
+ vars:
112
+ - [ MET_pt, null ]
113
+ - [ MET_phi, null ]
114
+ - [ scouting_trig, null]
115
+ - [ offline_trig, null]
116
+ - [ veto_trig, null ]
117
+ n_electrons:
118
+ pad_mode: wrap
119
+ length: 1
120
+ vars:
121
+ - [ nElectron, null ]
122
+ n_photons:
123
+ pad_mode: wrap
124
+ length: 1
125
+ vars:
126
+ - [ nPhotons, null ]
127
+ n_muons:
128
+ pad_mode: wrap
129
+ length: 1
130
+ vars:
131
+ - [ nMuons, null ]
132
+ electrons:
133
+ pad_mode: wrap
134
+ length: 10
135
+ vars:
136
+ - [ Electron_pt, null ]
137
+ - [ Electron_eta, null ]
138
+ - [ Electron_phi, null ]
139
+ - [ Electron_charge, null ]
140
+ muons:
141
+ pad_mode: wrap
142
+ length: 10
143
+ vars:
144
+ - [ Muon_pt, null ]
145
+ - [ Muon_eta, null ]
146
+ - [ Muon_phi, null ]
147
+ - [ Muon_charge, null ]
148
+ photons:
149
+ pad_mode: wrap
150
+ length: 10
151
+ vars:
152
+ - [ Photon_pt, null ]
153
+ - [ Photon_eta, null ]
154
+ - [ Photon_phi, null ]
155
+ matrix_element_gen_particles:
156
+ pad_mode: wrap
157
+ length: 2
158
+ vars:
159
+ - [MatrixElementGenParticle_pt, null]
160
+ - [MatrixElementGenParticle_eta, null]
161
+ - [MatrixElementGenParticle_phi, null]
162
+ - [MatrixElementGenParticle_mass, null]
163
+ - [MatrixElementGenParticle_pdgId, null]
164
+ labels:
165
+
166
+ observers:
167
+ #- recojet_e
168
+ #- recojet_theta
169
+ #- recojet_phi
170
+ #- recojet_m
171
+ #- n_pfcand
172
+
config_files/config_jets_1.yaml ADDED
@@ -0,0 +1,191 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ treename: null
2
+ selection:
3
+ ### use `&`, `|`, `~` for logical operations on numpy arrays
4
+ ### can use functions from `math`, `np` (numpy), and `awkward` in the expression
5
+ #(jet_tightId==1) & (jet_no<2) & (fj_pt>200) & (fj_pt<2500) & (((sample_isQCD==0) & (fj_isQCD==0)) | ((sample_isQCD==1) & (fj_isQCD==1))) & (event_no%7!=0)
6
+ #(recojet_e>=5)
7
+
8
+ test_time_selection:
9
+ ### selection to apply at test time (i.e., when running w/ --predict)
10
+ #(jet_tightId==1) & (jet_no<2) & (fj_pt>200) & (fj_pt<2500) & (((sample_isQCD==0) & (fj_isQCD==0)) | ((sample_isQCD==1) & (fj_isQCD==1))) & (event_no%7==0)
11
+ #(recojet_e<5)
12
+
13
+ new_variables:
14
+ ### [format] name: formula
15
+ ### can use functions from `math`, `np` (numpy), and `awkward` in the expression
16
+ #pfcand_mask: awkward.JaggedArray.ones_like(pfcand_etarel)
17
+ #sv_mask: awkward.JaggedArray.ones_like(sv_etarel)
18
+ #pfcand_mask: awkward.JaggedArray.ones_like(pfcand_e)
19
+
20
+ preprocess:
21
+ ### method: [manual, auto] - whether to use manually specified parameters for variable standardization
22
+ ### [note]: `[var]_mask` will not be transformed even if `method=auto`
23
+
24
+ inputs:
25
+ n_fat_jets:
26
+ pad_mode: wrap
27
+ length: 1
28
+ vars:
29
+ - [nFatJet, null]
30
+ # - [nJetId, null]
31
+ fat_jets:
32
+ pad_mode: wrap
33
+ length: 50
34
+ vars:
35
+ - [FatJet_pt, null]
36
+ - [FatJet_eta, null]
37
+ - [FatJet_phi, null]
38
+ - [FatJet_mass, null]
39
+ n_jets:
40
+ pad_mode: wrap
41
+ length: 1
42
+ vars:
43
+ - [ nJet, null ]
44
+ jets:
45
+ pad_mode: wrap
46
+ length: 50
47
+ vars:
48
+ - [ Jet_pt, null ]
49
+ - [ Jet_eta, null ]
50
+ - [ Jet_phi, null ]
51
+ - [ Jet_mass, null ]
52
+ n_genjets:
53
+ pad_mode: wrap
54
+ length: 1
55
+ vars:
56
+ - [n_genjet, null]
57
+ genjets:
58
+ pad_mode: wrap
59
+ length: 50
60
+ vars:
61
+ - [GenFatJet_pt, null]
62
+ - [GenFatJet_eta, null]
63
+ - [GenFatJet_phi, null]
64
+ - [GenFatJet_mass, null]
65
+ n_pfcands:
66
+ pad_mode: wrap
67
+ length: 1
68
+ vars:
69
+ - [ nPFCands, null ]
70
+ pfcands:
71
+ pad_mode: wrap
72
+ length: 750
73
+ vars:
74
+ - [PFCands_pt, null]
75
+ - [PFCands_eta, null]
76
+ - [PFCands_phi, null]
77
+ - [PFCands_mass, null]
78
+ - [PFCands_charge, null]
79
+ - [PFCands_pdgId, null]
80
+
81
+ pfcands_jet_mapping:
82
+ pad_mode: wrap
83
+ length: 750
84
+ vars:
85
+ - [ FatJetPFCands_jetIdx, null ]
86
+ - [ FatJetPFCands_pFCandsIdx, null ]
87
+ #n_offline_pfcands:
88
+ # pad_mode: wrap
89
+ # length: 1
90
+ # vars:
91
+ # - [ nOfflinePFCands, null ]
92
+ #offline_pfcands:
93
+ # pad_mode: wrap
94
+ # length: 750
95
+ # vars:
96
+ # - [ OfflinePFCands_pt, null ]
97
+ # - [ OfflinePFCands_eta, null ]
98
+ # - [ OfflinePFCands_phi, null ]
99
+ # - [ OfflinePFCands_mass, null ]
100
+ # - [ OfflinePFCands_charge, null ]
101
+ # - [ OfflinePFCands_pdgId, null ]
102
+ #offline_pfcands_jet_mapping:
103
+ # pad_mode: wrap
104
+ # length: 750
105
+ # vars:
106
+ # - [ OfflineFatJetPFCands_jetIdx, null ]
107
+ # - [ OfflineFatJetPFCands_pFCandsIdx, null ]
108
+ MET:
109
+ pad_mode: wrap
110
+ length: 1
111
+ vars:
112
+ - [ MET_pt, null ]
113
+ - [ MET_phi, null ]
114
+ - [ scouting_trig, null]
115
+ - [ offline_trig, null]
116
+ - [ veto_trig, null ]
117
+ n_electrons:
118
+ pad_mode: wrap
119
+ length: 1
120
+ vars:
121
+ - [ nElectron, null ]
122
+ n_photons:
123
+ pad_mode: wrap
124
+ length: 1
125
+ vars:
126
+ - [ nPhotons, null ]
127
+ n_muons:
128
+ pad_mode: wrap
129
+ length: 1
130
+ vars:
131
+ - [ nMuons, null ]
132
+ electrons:
133
+ pad_mode: wrap
134
+ length: 10
135
+ vars:
136
+ - [ Electron_pt, null ]
137
+ - [ Electron_eta, null ]
138
+ - [ Electron_phi, null ]
139
+ - [ Electron_charge, null ]
140
+ muons:
141
+ pad_mode: wrap
142
+ length: 10
143
+ vars:
144
+ - [ Muon_pt, null ]
145
+ - [ Muon_eta, null ]
146
+ - [ Muon_phi, null ]
147
+ - [ Muon_charge, null ]
148
+ photons:
149
+ pad_mode: wrap
150
+ length: 10
151
+ vars:
152
+ - [ Photon_pt, null ]
153
+ - [ Photon_eta, null ]
154
+ - [ Photon_phi, null ]
155
+ matrix_element_gen_particles:
156
+ pad_mode: wrap
157
+ length: 2
158
+ vars:
159
+ - [MatrixElementGenParticle_pt, null]
160
+ - [MatrixElementGenParticle_eta, null]
161
+ - [MatrixElementGenParticle_phi, null]
162
+ - [MatrixElementGenParticle_mass, null]
163
+ - [MatrixElementGenParticle_pdgId, null]
164
+ final_gen_particles:
165
+ pad_mode: wrap
166
+ length: 2000
167
+ vars:
168
+ - [FinalGenParticle_pt, null]
169
+ - [FinalGenParticle_eta, null]
170
+ - [FinalGenParticle_phi, null]
171
+ - [FinalGenParticle_mass, null]
172
+ - [FinalGenParticle_pdgId, null]
173
+ - [FinalGenParticle_status, null]
174
+ final_parton_level_particles:
175
+ pad_mode: wrap
176
+ length: 400
177
+ vars:
178
+ - [FinalPartonLevelParticle_pt, null]
179
+ - [FinalPartonLevelParticle_eta, null]
180
+ - [FinalPartonLevelParticle_phi, null]
181
+ - [FinalPartonLevelParticle_mass, null]
182
+ - [FinalPartonLevelParticle_pdgId, null]
183
+ - [FinalPartonLevelParticle_status, null]
184
+
185
+ observers:
186
+ #- recojet_e
187
+ #- recojet_theta
188
+ #- recojet_phi
189
+ #- recojet_m
190
+ #- n_pfcand
191
+
config_files/config_jets_1_delphes.yaml ADDED
@@ -0,0 +1,86 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ treename: Delphes;1
2
+ selection:
3
+ ### use `&`, `|`, `~` for logical operations on numpy arrays
4
+ ### can use functions from `math`, `np` (numpy), and `awkward` in the expression
5
+ #(jet_tightId==1) & (jet_no<2) & (fj_pt>200) & (fj_pt<2500) & (((sample_isQCD==0) & (fj_isQCD==0)) | ((sample_isQCD==1) & (fj_isQCD==1))) & (event_no%7!=0)
6
+ #(recojet_e>=5)
7
+
8
+ test_time_selection:
9
+ ### selection to apply at test time (i.e., when running w/ --predict)
10
+ #(jet_tightId==1) & (jet_no<2) & (fj_pt>200) & (fj_pt<2500) & (((sample_isQCD==0) & (fj_isQCD==0)) | ((sample_isQCD==1) & (fj_isQCD==1))) & (event_no%7==0)
11
+ #(recojet_e<5)
12
+
13
+ new_variables:
14
+ ### [format] name: formula
15
+ ### can use functions from `math`, `np` (numpy), and `awkward` in the expression
16
+ #pfcand_mask: awkward.JaggedArray.ones_like(pfcand_etarel)
17
+ #sv_mask: awkward.JaggedArray.ones_like(sv_etarel)
18
+ #pfcand_mask: awkward.JaggedArray.ones_like(pfcand_e)
19
+
20
+ preprocess:
21
+ ### method: [manual, auto] - whether to use manually specified parameters for variable standardization
22
+ ### [note]: `[var]_mask` will not be transformed even if `method=auto`
23
+
24
+ inputs:
25
+ n_CH:
26
+ pad_mode: wrap
27
+ length: 1
28
+ vars:
29
+ - [ EFlowTrack_size, null ]
30
+ n_NH:
31
+ pad_mode: wrap
32
+ length: 1
33
+ vars:
34
+ - [ EFlowNeutralHadron_size, null ]
35
+ n_photon:
36
+ pad_mode: wrap
37
+ length: 1
38
+ vars:
39
+ - [ EFlowPhoton_size, null ]
40
+ CH:
41
+ pad_mode: wrap
42
+ length: 1500
43
+ vars:
44
+ - [EFlowTrack.Eta, null]
45
+ - [EFlowTrack.Phi, null]
46
+ - [EFlowTrack.PT, null]
47
+ - [EFlowTrack.Mass, null]
48
+ - [EFlowTrack.Charge, null]
49
+ NH:
50
+ pad_mode: wrap
51
+ length: 1500
52
+ vars:
53
+ - [EFlowNeutralHadron.Eta, null]
54
+ - [EFlowNeutralHadron.Phi, null]
55
+ - [EFlowNeutralHadron.ET, null]
56
+ EFlowPhoton:
57
+ pad_mode: wrap
58
+ length: 1500
59
+ vars:
60
+ - [EFlowPhoton.Eta, null]
61
+ - [EFlowPhoton.Phi, null]
62
+ - [EFlowPhoton.ET, null]
63
+ GenParticles:
64
+ pad_mode: wrap
65
+ length: 7500
66
+ vars:
67
+ - [Particle.Eta, null]
68
+ - [Particle.Phi, null]
69
+ - [Particle.PT, null]
70
+ - [Particle.Charge, null]
71
+ - [Particle.Mass, null]
72
+ - [Particle.PID, null]
73
+ - [Particle.Status, null]
74
+ NParticles:
75
+ pad_mode: wrap
76
+ length: 1
77
+ vars:
78
+ - [Particle_size, null]
79
+ observers:
80
+ #- recojet_e
81
+ #- recojet_theta
82
+ #- recojet_phi
83
+ #- recojet_m
84
+ #- n_pfcand
85
+
86
+
config_files/config_jets_2_delphes.yaml ADDED
@@ -0,0 +1,63 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ treename: Delphes;1
2
+ selection:
3
+ ### use `&`, `|`, `~` for logical operations on numpy arrays
4
+ ### can use functions from `math`, `np` (numpy), and `awkward` in the expression
5
+ #(jet_tightId==1) & (jet_no<2) & (fj_pt>200) & (fj_pt<2500) & (((sample_isQCD==0) & (fj_isQCD==0)) | ((sample_isQCD==1) & (fj_isQCD==1))) & (event_no%7!=0)
6
+ #(recojet_e>=5)
7
+
8
+ test_time_selection:
9
+ ### selection to apply at test time (i.e., when running w/ --predict)
10
+ #(jet_tightId==1) & (jet_no<2) & (fj_pt>200) & (fj_pt<2500) & (((sample_isQCD==0) & (fj_isQCD==0)) | ((sample_isQCD==1) & (fj_isQCD==1))) & (event_no%7==0)
11
+ #(recojet_e<5)
12
+
13
+ new_variables:
14
+ ### [format] name: formula
15
+ ### can use functions from `math`, `np` (numpy), and `awkward` in the expression
16
+ #pfcand_mask: awkward.JaggedArray.ones_like(pfcand_etarel)
17
+ #sv_mask: awkward.JaggedArray.ones_like(sv_etarel)
18
+ #pfcand_mask: awkward.JaggedArray.ones_like(pfcand_e)
19
+
20
+ preprocess:
21
+ ### method: [manual, auto] - whether to use manually specified parameters for variable standardization
22
+ ### [note]: `[var]_mask` will not be transformed even if `method=auto`
23
+
24
+ inputs:
25
+ n_PFCands:
26
+ pad_mode: wrap
27
+ length: 1
28
+ vars:
29
+ - [ ParticleFlowCandidate_size, null ]
30
+ PFCands:
31
+ pad_mode: wrap
32
+ length: 1500
33
+ vars:
34
+ - [ParticleFlowCandidate.Eta, null]
35
+ - [ParticleFlowCandidate.Phi, null]
36
+ - [ParticleFlowCandidate.PT, null]
37
+ - [ParticleFlowCandidate.Mass, null]
38
+ - [ParticleFlowCandidate.Charge, null]
39
+ - [ParticleFlowCandidate.PID, null]
40
+ GenParticles:
41
+ pad_mode: wrap
42
+ length: 7500
43
+ vars:
44
+ - [Particle.Eta, null]
45
+ - [Particle.Phi, null]
46
+ - [Particle.PT, null]
47
+ - [Particle.Charge, null]
48
+ - [Particle.Mass, null]
49
+ - [Particle.PID, null]
50
+ - [Particle.Status, null]
51
+ NParticles:
52
+ pad_mode: wrap
53
+ length: 1
54
+ vars:
55
+ - [Particle_size, null]
56
+ observers:
57
+ #- recojet_e
58
+ #- recojet_theta
59
+ #- recojet_phi
60
+ #- recojet_m
61
+ #- n_pfcand
62
+
63
+
container_shell.sh ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
2
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
3
+ singularity shell -B /work/gkrzmanc/ --nv docker://dologarcia/gatr:v0
4
+
docker-compose.yaml ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ version: '3'
2
+
3
+ services:
4
+ app:
5
+ image: gkrz/jetclustering_demo:v0
6
+ ports:
7
+ - "7860:7860"
env.sh ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # For CERN machines
2
+ #export SVJ_CODE_ROOT="/eos/home-g/gkrzmanc/jetclustering/code"
3
+ #export SVJ_DATA_ROOT="/eos/home-g/gkrzmanc/jetclustering/data"
4
+ #export SVJ_RESULTS_ROOT="/eos/home-g/gkrzmanc/jetclustering/results"
5
+ #export SVJ_PREPROCESSED_DATA_ROOT="/eos/home-g/gkrzmanc/jetclustering/preprocessed_data"
6
+
7
+
8
+ # For PSI T3
9
+ export SVJ_CODE_ROOT="/work/gkrzmanc/jetclustering/code"
10
+ #export SVJ_DATA_ROOT="/work/gkrzmanc/jetclustering/data"
11
+ export SVJ_DATA_ROOT="/pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc/jetclustering/data"
12
+ export SVJ_RESULTS_ROOT="/work/gkrzmanc/jetclustering/results"
13
+ export SVJ_PREPROCESSED_DATA_ROOT="/work/gkrzmanc/jetclustering/preprocessed_data"
14
+ #export SVJ_PREPROCESSED_DATA_ROOT="/pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc/jetclustering/preprocessed_data"
15
+ export SVJ_WANDB_ENTITY="fcc_ml"
16
+ export SVJ_RESULTS_ROOT_FALLBACK="/pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc/jetclustering/results"
17
+
jobs/BigTraining_2_spatial_part_only_t3.slurm ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ #SBATCH --partition=qgpu # Specify the partition
3
+ #SBATCH --account=gpu_gres # Specify the account
4
+ #SBATCH --mem=3000 # Request 10GB of memory
5
+ #SBATCH --time=00:10:00
6
+ #SBATCH --job-name=SVJtr3 # Name the job
7
+ #SBATCH --output=jobs/BigTraining_output.log # Redirect stdout to a log file
8
+ #SBATCH --error=jobs/BigTraining_error.log # Redirect stderr to a log file
9
+ #SBATCH --gres=gpu:1
10
+ source env.sh
11
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
12
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
13
+ nvidia-smi
14
+
15
+ srun singularity exec -B /t3home/gkrzmanc/ -B /work/gkrzmanc/ --nv docker://gkrz/lgatr:v3 python -m src.train -train scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1100_mDark-20_rinv-0.3_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1500_mDark-20_rinv-0.3_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1100_mDark-20_rinv-0.5_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1500_mDark-20_rinv-0.5_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1100_mDark-20_rinv-0.7_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1500_mDark-20_rinv-0.7_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1200_mDark-20_rinv-0.3_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-700_mDark-20_rinv-0.3_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1200_mDark-20_rinv-0.5_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-700_mDark-20_rinv-0.5_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1200_mDark-20_rinv-0.7_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-700_mDark-20_rinv-0.7_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1300_mDark-20_rinv-0.3_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-800_mDark-20_rinv-0.3_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1300_mDark-20_rinv-0.5_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-800_mDark-20_rinv-0.5_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1300_mDark-20_rinv-0.7_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-800_mDark-20_rinv-0.7_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1400_mDark-20_rinv-0.3_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-900_mDark-20_rinv-0.3_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1400_mDark-20_rinv-0.5_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-900_mDark-20_rinv-0.5_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1400_mDark-20_rinv-0.7_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-900_mDark-20_rinv-0.7_alpha-peak_13TeV-pythia8_n-2000 -val scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-1000_mDark-20_rinv-0.3 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-1500_mDark-20_rinv-0.5 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-700_mDark-20_rinv-0.7 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-900_mDark-20_rinv-0.3 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-1000_mDark-20_rinv-0.5 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-1500_mDark-20_rinv-0.7 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-800_mDark-20_rinv-0.3 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-900_mDark-20_rinv-0.5 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-1000_mDark-20_rinv-0.7 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-700_mDark-20_rinv-0.3 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-800_mDark-20_rinv-0.5 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-900_mDark-20_rinv-0.7 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-1500_mDark-20_rinv-0.3 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-700_mDark-20_rinv-0.5 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-800_mDark-20_rinv-0.7 -net src/models/LGATr/lgatr.py -bs 64 --gpus 0 --run-name Train_LGATr_SB_spatial_part_only --val-dataset-size 4000 --num-epochs 1000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --spatial-part-only --validation-steps 2000 --num-workers 0
jobs/BigTraining_2_spatial_part_only_vega.slurm ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ #SBATCH --job-name="SVJtrAll"
3
+ #SBATCH --time=48:00:00
4
+ #SBATCH --nodes=1
5
+ #SBATCH --gres=gpu:1
6
+ #SBATCH --ntasks-per-core=1
7
+ #SBATCH --ntasks-per-node=1
8
+ #SBATCH --cpus-per-task=2
9
+ #SBATCH --partition=gpu
10
+ #SBATCH --mem=25GB
11
+ #SBATCH --output=jobs/big_training_2_output1.log
12
+ #SBATCH --error=jobs/big_training_2_error1.log
13
+
14
+
15
+ source env.sh
16
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
17
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
18
+ nvidia-smi
19
+ srun singularity exec -B /ceph/hpc/home/krzmancg --nv docker://gkrz/lgatr:v3 python -m src.train -train scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1100_mDark-20_rinv-0.3_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1500_mDark-20_rinv-0.3_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1100_mDark-20_rinv-0.5_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1500_mDark-20_rinv-0.5_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1100_mDark-20_rinv-0.7_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1500_mDark-20_rinv-0.7_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1200_mDark-20_rinv-0.3_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-700_mDark-20_rinv-0.3_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1200_mDark-20_rinv-0.5_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-700_mDark-20_rinv-0.5_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1200_mDark-20_rinv-0.7_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-700_mDark-20_rinv-0.7_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1300_mDark-20_rinv-0.3_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-800_mDark-20_rinv-0.3_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1300_mDark-20_rinv-0.5_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-800_mDark-20_rinv-0.5_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1300_mDark-20_rinv-0.7_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-800_mDark-20_rinv-0.7_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1400_mDark-20_rinv-0.3_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-900_mDark-20_rinv-0.3_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1400_mDark-20_rinv-0.5_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-900_mDark-20_rinv-0.5_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-1400_mDark-20_rinv-0.7_alpha-peak_13TeV-pythia8_n-2000 scouting_PFNano_signals2/SVJ_hadronic_std3/s-channel_mMed-900_mDark-20_rinv-0.7_alpha-peak_13TeV-pythia8_n-2000 -val scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-1000_mDark-20_rinv-0.3 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-1500_mDark-20_rinv-0.5 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-700_mDark-20_rinv-0.7 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-900_mDark-20_rinv-0.3 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-1000_mDark-20_rinv-0.5 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-1500_mDark-20_rinv-0.7 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-800_mDark-20_rinv-0.3 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-900_mDark-20_rinv-0.5 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-1000_mDark-20_rinv-0.7 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-700_mDark-20_rinv-0.3 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-800_mDark-20_rinv-0.5 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-900_mDark-20_rinv-0.7 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-1500_mDark-20_rinv-0.3 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-700_mDark-20_rinv-0.5 scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-800_mDark-20_rinv-0.7 -net src/models/LGATr/lgatr.py -bs 256 --gpus 0 --run-name Train_LGATr_SB_All_data_CONT --val-dataset-size 4000 --num-epochs 1000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --spatial-part-only --validation-steps 1000 --load-model-weights train/Train_LGATr_SB_All_data_2025_01_14_13_19_34/step_5000_epoch_1.ckpt
20
+
21
+ # sbatch jobs/BigTraining_2_spatial_part_only_vega.slurm
jobs/IRC_training/Delphes_training_t3_NoPID_augment.sh ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesTrainSVJ # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/Dtr_lgatr_aug.out
11
+ #SBATCH --error=jobs/vega/Dtr_lgatr_aug.err
12
+
13
+ source env.sh
14
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
15
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
16
+ nvidia-smi
17
+
18
+ srun singularity exec -B /t3home/gkrzmanc/ -B /work/gkrzmanc/ -B /pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train2_PU_PFfix_part9/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-700_mDark-20_rinv-0.7 -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name LGATr_Aug --val-dataset-size 150 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius 0.8 --num-blocks 10 -mv-ch 16 -s-ch 64 --spatial-part-only --validation-steps 60 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_10_16_64_0.8_2025_05_03_18_35_53_134/step_70000_epoch_16.ckpt --num-workers 0
19
+
20
+ exit 0
21
+ EOT
22
+
23
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
24
+ # bash jobs/IRC_training/Delphes_training_t3_NoPID_augment.sh
jobs/IRC_training/Delphes_training_t3_NoPID_augment_IRC.sh ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesTrainSVJ # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/Dtr_lgatr_aug.out
11
+ #SBATCH --error=jobs/vega/Dtr_lgatr_aug.err
12
+
13
+ source env.sh
14
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
15
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
16
+ nvidia-smi
17
+ srun singularity exec -B /t3home/gkrzmanc/ -B /work/gkrzmanc/ -B /pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train2_PU_PFfix_part9/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name Delphes_Aug_IRCSplit_CONT --val-dataset-size 150 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius 0.8 --num-blocks 10 -mv-ch 16 -s-ch 64 --spatial-part-only --validation-steps 60 --no-pid --augment-soft-particles --load-model-weights train/Delphes_Aug_IRCSplit_2025_05_06_10_09_00_567/step_8820_epoch_1.ckpt --num-workers 0 -irc
18
+
19
+
20
+ exit 0
21
+ EOT
22
+
23
+ # bash jobs/IRC_training/Delphes_training_t3_NoPID_augment_IRC.sh
24
+
jobs/IRC_training/Delphes_training_t3_NoPID_augment_IRC_SN.sh ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesTrainSVJ # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/Dtr_lgatr_aug.out
11
+ #SBATCH --error=jobs/vega/Dtr_lgatr_aug.err
12
+
13
+ source env.sh
14
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
15
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
16
+ nvidia-smi
17
+ srun singularity exec -B /t3home/gkrzmanc/ -B /work/gkrzmanc/ -B /pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train2_PU_PFfix_part9/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name Delphes_Aug_IRC_Split_and_Noise_CONT --val-dataset-size 150 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius 0.8 --num-blocks 10 -mv-ch 16 -s-ch 64 --spatial-part-only --validation-steps 60 --no-pid --augment-soft-particles --num-workers 0 -irc --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_10_16_64_0.8_2025_05_03_18_35_53_134/step_70000_epoch_16.ckpt
18
+
19
+
20
+ exit 0
21
+ EOT
22
+
23
+ # bash jobs/IRC_training/Delphes_training_t3_NoPID_augment_IRC_SN.sh
24
+
jobs/IRC_training/start_at_30k/Delphes_training_t3_NoPID_augment.sh ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesTrainSVJ # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/Dtr_lgatr_aug.out
11
+ #SBATCH --error=jobs/vega/Dtr_lgatr_aug.err
12
+
13
+ source env.sh
14
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
15
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
16
+ nvidia-smi
17
+
18
+ srun singularity exec -B /t3home/gkrzmanc/ -B /work/gkrzmanc/ -B /pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train2_PU_PFfix_part9/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-700_mDark-20_rinv-0.7 -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name LGATr_Aug_30k --val-dataset-size 150 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius 0.8 --num-blocks 10 -mv-ch 16 -s-ch 64 --spatial-part-only --validation-steps 60 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_10_16_64_0.8_2025_05_03_18_35_53_134/step_30000_epoch_7.ckpt --num-workers 0
19
+
20
+ exit 0
21
+ EOT
22
+
23
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
24
+ # bash jobs/IRC_training/start_at_30k/Delphes_training_t3_NoPID_augment.sh
jobs/IRC_training/start_at_30k/Delphes_training_t3_NoPID_augment_IRC.sh ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesTrainSVJ # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/Dtr_lgatr_aug.out
11
+ #SBATCH --error=jobs/vega/Dtr_lgatr_aug.err
12
+
13
+ source env.sh
14
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
15
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
16
+ nvidia-smi
17
+
18
+
19
+ srun singularity exec -B /t3home/gkrzmanc/ -B /work/gkrzmanc/ -B /pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train2_PU_PFfix_part9/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name Delphes_Aug_IRCSplit_30k_from10k --val-dataset-size 150 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius 0.8 --num-blocks 10 -mv-ch 16 -s-ch 64 --spatial-part-only --validation-steps 60 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_10_16_64_0.8_2025_05_03_18_35_53_134/step_30000_epoch_7.ckpt --num-workers 0 -irc
20
+
21
+
22
+ exit 0
23
+ EOT
24
+
25
+ # bash jobs/IRC_training/start_at_30k/Delphes_training_t3_NoPID_augment_IRC.sh
26
+ ####
jobs/IRC_training/start_at_50k/Delphes_training_t3_NoPID_augment.sh ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesTrainSVJ # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/Dtr_lgatr_aug.out
11
+ #SBATCH --error=jobs/vega/Dtr_lgatr_aug.err
12
+
13
+ source env.sh
14
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
15
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
16
+ nvidia-smi
17
+
18
+ srun singularity exec -B /t3home/gkrzmanc/ -B /work/gkrzmanc/ -B /pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train2_PU_PFfix_part9/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val scouting_PFNano_signals2/SVJ_hadronic_std2/s-channel_mMed-700_mDark-20_rinv-0.7 -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name LGATr_Aug_50k --val-dataset-size 150 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius 0.8 --num-blocks 10 -mv-ch 16 -s-ch 64 --spatial-part-only --validation-steps 60 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_10_16_64_0.8_2025_05_03_18_35_53_134/step_50000_epoch_12.ckpt --num-workers 0
19
+
20
+ exit 0
21
+ EOT
22
+
23
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
24
+ # bash jobs/IRC_training/start_at_50k/Delphes_training_t3_NoPID_augment.sh
jobs/IRC_training/start_at_50k/Delphes_training_t3_NoPID_augment_IRC.sh ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesTrainSVJ # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/Dtr_lgatr_aug.out
11
+ #SBATCH --error=jobs/vega/Dtr_lgatr_aug.err
12
+
13
+ source env.sh
14
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
15
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
16
+ nvidia-smi
17
+ #srun singularity exec -B /t3home/gkrzmanc/ -B /work/gkrzmanc/ -B /pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train2_PU_PFfix_part9/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name Delphes_Aug_IRCSplit_50k --val-dataset-size 150 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius 0.8 --num-blocks 10 -mv-ch 16 -s-ch 64 --spatial-part-only --validation-steps 60 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_10_16_64_0.8_2025_05_03_18_35_53_134/step_50000_epoch_12.ckpt --num-workers 0 -irc
18
+ srun singularity exec -B /t3home/gkrzmanc/ -B /work/gkrzmanc/ -B /pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train2_PU_PFfix_part9/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name Delphes_Aug_IRCSplit_50k_from10k --val-dataset-size 150 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius 0.8 --num-blocks 10 -mv-ch 16 -s-ch 64 --spatial-part-only --validation-steps 60 --no-pid --augment-soft-particles --load-model-weights train/Delphes_Aug_IRCSplit_50k_2025_05_09_15_22_38_956/step_13620_epoch_2.ckpt --num-workers 0 -irc
19
+
20
+ -irc
21
+ exit 0
22
+ EOT
23
+
24
+ # bash jobs/IRC_training/start_at_50k/Delphes_training_t3_NoPID_augment_IRC.sh
25
+ ####
jobs/IRC_training/start_at_50k/Delphes_training_t3_NoPID_augment_IRCSN.sh ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesTrainSVJ # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/Dtr_lgatr_aug.out
11
+ #SBATCH --error=jobs/vega/Dtr_lgatr_aug.err
12
+
13
+ source env.sh
14
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
15
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
16
+ nvidia-smi
17
+
18
+ srun singularity exec -B /t3home/gkrzmanc/ -B /work/gkrzmanc/ -B /pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train2_PU_PFfix_part9/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name Delphes_Aug_IRCSplit_50k_SN_from3kFT --val-dataset-size 150 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius 0.8 --num-blocks 10 -mv-ch 16 -s-ch 64 --spatial-part-only --validation-steps 60 --no-pid --augment-soft-particles --load-model-weights train/Delphes_Aug_IRCSplit_50k_SN_2025_05_12_13_57_45_477/step_3060_epoch_1.ckpt --num-workers 0 -irc
19
+
20
+
21
+ exit 0
22
+ EOT
23
+
24
+ # bash jobs/IRC_training/start_at_50k/Delphes_training_t3_NoPID_augment_IRCSN.sh
25
+ ####
jobs/IRC_training/start_at_50k/Delphes_training_t3_NoPID_augment_IRC_noaug.sh ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesTrainSVJ # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/Dtr_lgatr_augGPU.out
11
+ #SBATCH --error=jobs/vega/Dtr_lgatr_augGPU.err
12
+
13
+ source env.sh
14
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
15
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
16
+ export PATH=/t3home/gkrzmanc/.local/lib/site-packages:$PATH
17
+ nvidia-smi
18
+ env
19
+ echo " ---- end env ---- "
20
+
21
+ srun singularity exec -B /work/gkrzmanc/ -B /pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc -B /t3home/gkrzmanc -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -c "import fastjet"
22
+ echo "Hello"
23
+ srun singularity exec -B /work/gkrzmanc/ -B /pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc -B /t3home/gkrzmanc -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train2_PU_PFfix_part9/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name Delphes_NOAug_IRCSplit_50k_cont --val-dataset-size 150 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius 0.8 --num-blocks 10 -mv-ch 16 -s-ch 64 --spatial-part-only --validation-steps 60 --no-pid --load-model-weights train/Delphes_NOAug_IRCSplit_50k__2025_05_13_09_56_39_345/step_2460_epoch_1.ckpt --num-workers 0 -irc
24
+
25
+ exit 0
26
+ EOT
27
+
28
+
29
+
30
+ # bash jobs/IRC_training/start_at_50k/Delphes_training_t3_NoPID_augment_IRC_noaug.sh
31
+ ####
32
+
33
+
34
+
jobs/IRC_training/start_at_50k/test.sh ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=short # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=00:01:00
7
+ #SBATCH --job-name=test1 # Name the job
8
+ #SBATCH --account=t3 # Specify the account
9
+ #SBATCH --output=jobs/vega/Dtr_lgatr_aug.out
10
+ #SBATCH --error=jobs/vega/Dtr_lgatr_aug.err
11
+
12
+ source env.sh
13
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
14
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
15
+
16
+ nvidia-smi
17
+ env
18
+ echo " ---- end env ---- "
19
+ srun singularity exec -B /t3home/gkrzmanc/ -B /work/gkrzmanc/ -B /pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m pip install --no-input fastjet
20
+
21
+ srun singularity exec -B /t3home/gkrzmanc/ -B /work/gkrzmanc/ -B /pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -c "import fastjet"
22
+ echo "Hello"
23
+ srun singularity exec -B /t3home/gkrzmanc/ -B /work/gkrzmanc/ -B /pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train2_PU_PFfix_part9/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name Delphes_NOAug_IRCSplit_50k_cont --val-dataset-size 150 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius 0.8 --num-blocks 10 -mv-ch 16 -s-ch 64 --spatial-part-only --validation-steps 60 --no-pid --load-model-weights train/Delphes_NOAug_IRCSplit_50k__2025_05_13_09_56_39_345/step_2460_epoch_1.ckpt --num-workers 0 -irc
24
+
25
+ exit 0
26
+ EOT
27
+
28
+ # bash jobs/IRC_training/start_at_50k/test.sh
29
+ ####
30
+
31
+
32
+
jobs/base_training/gatr_training_NoPIDDelphes.sh ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesSVJTrain # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/DTrGATr_out_$1_$2_$3_R$4.log
11
+ #SBATCH --error=jobs/vega/DTrGATr_err_$1_$2_$3_R$4.log
12
+
13
+ source env.sh
14
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
15
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
16
+ nvidia-smi
17
+
18
+ srun singularity exec -B /work -B /pnfs -B /t3home --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train2_PU_PFfix_part9/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -net src/models/GATr/Gatr.py -bs 20 --gpus 0 --run-name GATr_training_NoPID_Delphes_PU_CoordFix_CONT_$1_$2_$3_$4 --val-dataset-size 1000 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius $4 --num-blocks $1 -mv-ch $2 -s-ch $3 --spatial-part-only --validation-steps 2000 --no-pid --load-model-weights train/GATr_training_NoPID_Delphes_PU_CoordFix_10_16_64_0.8_2025_05_05_13_06_27_898/step_50000_epoch_12.ckpt
19
+
20
+ exit 0
21
+ EOT
22
+
23
+
24
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
25
+ # bash jobs/base_training/gatr_training_NoPIDDelphes.sh 10 16 64 0.8
26
+
jobs/base_training/lgatr_training_NoPIDDelphes.sh ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesSVJTrain # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/DTrLGATr_out_$1_$2_$3_R$4.log
11
+ #SBATCH --error=jobs/vega/DTrLGATr_err_$1_$2_$3_R$4.log
12
+
13
+
14
+ source env.sh
15
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
16
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
17
+ nvidia-smi
18
+
19
+ srun singularity exec -B /work -B /pnfs -B /t3home --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train2_PU_PFfix_part9/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -net src/models/LGATr/lgatr.py -bs 20 --gpus 0 --run-name LGATr_training_NoPID_Delphes_PU_PFfix_$1_$2_$3_$4 --val-dataset-size 1000 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius $4 --num-blocks $1 -mv-ch $2 -s-ch $3 --spatial-part-only --validation-steps 2000 --no-pid
20
+
21
+ exit 0
22
+ EOT
23
+
24
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
25
+ # bash jobs/base_training/lgatr_700_07.sh 10 16 64 0.8
26
+
jobs/base_training/transformer_training_NoPIDDelphes.sh ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesSVJTrain # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/DTr_out_$1_$2_$3_R$4.log
11
+ #SBATCH --error=jobs/vega/DTr_err_$1_$2_$3_R$4.log
12
+
13
+ source env.sh
14
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
15
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
16
+ nvidia-smi
17
+
18
+ srun singularity exec -B /work -B /pnfs -B /t3home --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train2_PU_PFfix_part9/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -net src/models/transformer/transformer.py -bs 20 --gpus 0 --run-name Transformer_training_NoPID_Delphes_PU_CoordFix_$1_$2_$3_$4 --val-dataset-size 1000 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius $4 --num-blocks $1 -mv-ch $2 -s-ch $3 --spatial-part-only --validation-steps 2000 --no-pid
19
+
20
+ exit 0
21
+ EOT
22
+
23
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
24
+ # bash jobs/vega/transformer_training_NoPIDDelphes.sh 10 16 64 0.8
jobs/base_training_different_datasets/aug/lgatr_700_07.sh ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesSVJTrain # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/DTrLGATr_out_$1_$2_$3_R$4.log
11
+ #SBATCH --error=jobs/vega/DTrLGATr_err_$1_$2_$3_R$4.log
12
+
13
+
14
+ source env.sh
15
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
16
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
17
+ nvidia-smi
18
+
19
+ srun singularity exec -B /work -B /pnfs -B /t3home -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train3_PU_PFfix_part0/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part1/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part2/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part3/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part4/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part5/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part6/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part7/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part8/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak -val Delphes_020425_train3_PU_PFfix_part9/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak -net src/models/LGATr/lgatr.py -bs 20 --gpus 0 --run-name GP_LGATr_training_NoPID_Delphes_PU_PFfix_700_07_$1_$2_$3_$4 --val-dataset-size 1000 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius $4 --num-blocks $1 -mv-ch $2 -s-ch $3 --spatial-part-only --validation-steps 2000 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_700_07_10_16_64_0.8_2025_05_16_19_44_46_795/step_50000_epoch_12.ckpt
20
+
21
+ exit 0
22
+ EOT
23
+
24
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
25
+ # bash jobs/base_training_different_datasets/aug/lgatr_700_07.sh 10 16 64 0.8
26
+
jobs/base_training_different_datasets/aug/lgatr_700_07_and_900_03.sh ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesSVJTrain # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/DTrLGATr_out_$1_$2_$3_R$4.log
11
+ #SBATCH --error=jobs/vega/DTrLGATr_err_$1_$2_$3_R$4.log
12
+
13
+
14
+ source env.sh
15
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
16
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
17
+ nvidia-smi
18
+
19
+ srun singularity exec -B /work -B /pnfs -B /t3home -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train3_PU_PFfix_part0/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part1/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part2/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part3/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part4/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part5/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part6/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part7/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part8/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train3_PU_PFfix_part9/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak -net src/models/LGATr/lgatr.py -bs 20 --gpus 0 --run-name GP_LGATr_training_NoPID_Delphes_PU_PFfix_700_07_AND_900_03_$1_$2_$3_$4 --val-dataset-size 1000 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius $4 --num-blocks $1 -mv-ch $2 -s-ch $3 --spatial-part-only --validation-steps 2000 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_700_07_AND_900_03_10_16_64_0.8_2025_05_16_21_04_26_991/step_50000_epoch_6.ckpt
20
+
21
+ exit 0
22
+ EOT
23
+
24
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
25
+ # bash jobs/base_training_different_datasets/aug/lgatr_700_07_and_900_03.sh 10 16 64 0.8
jobs/base_training_different_datasets/aug/lgatr_700_07_and_900_03_and_QCD.sh ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesSVJTrain # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/DTrLGATr_out_$1_$2_$3_R$4.log
11
+ #SBATCH --error=jobs/vega/DTrLGATr_err_$1_$2_$3_R$4.log
12
+
13
+
14
+ source env.sh
15
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
16
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
17
+ nvidia-smi
18
+
19
+ srun singularity exec -B /work -B /pnfs -B /t3home -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train3_PU_PFfix_part0/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part1/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part2/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part3/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part4/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part5/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part6/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part7/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part8/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak QCDtrain_part0/qcd_test_0 QCDtrain_part0/qcd_test_1 QCDtrain_part0/qcd_test_2 QCDtrain_part0/qcd_test_3 QCDtrain_part0/qcd_test_4 QCDtrain_part0/qcd_test_5 QCDtrain_part0/qcd_test_6 QCDtrain_part0/qcd_test_7 QCDtrain_part0/qcd_test_8 -val Delphes_020425_train3_PU_PFfix_part9/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak -net src/models/LGATr/lgatr.py -bs 20 --gpus 0 --run-name GP_LGATr_training_NoPID_Delphes_PU_PFfix_700_07_AND_900_03_AND_QCD_$1_$2_$3_$4 --val-dataset-size 1000 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius $4 --num-blocks $1 -mv-ch $2 -s-ch $3 --spatial-part-only --validation-steps 2000 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_700_07_AND_900_03_AND_QCD_10_16_64_0.8_2025_05_16_21_04_26_937/step_50000_epoch_4.ckpt
20
+
21
+ exit 0
22
+ EOT
23
+
24
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
25
+ # bash jobs/base_training_different_datasets/aug/lgatr_700_07_and_900_03_and_QCD.sh 10 16 64 0.8
jobs/base_training_different_datasets/aug/lgatr_QCD.sh ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesSVJTrain # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/DTrLGATr_out_$1_$2_$3_R$4.log
11
+ #SBATCH --error=jobs/vega/DTrLGATr_err_$1_$2_$3_R$4.log
12
+
13
+
14
+ source env.sh
15
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
16
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
17
+ nvidia-smi
18
+
19
+ srun singularity exec -B /work -B /pnfs -B /t3home -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train QCDtrain_part0/qcd_test_0 QCDtrain_part0/qcd_test_1 QCDtrain_part0/qcd_test_2 QCDtrain_part0/qcd_test_3 QCDtrain_part0/qcd_test_4 QCDtrain_part0/qcd_test_5 QCDtrain_part0/qcd_test_6 QCDtrain_part0/qcd_test_7 QCDtrain_part0/qcd_test_8 -val QCDtrain_part0/qcd_test_9 -net src/models/LGATr/lgatr.py -bs 20 --gpus 0 --run-name GP_LGATr_training_NoPID_Delphes_PU_PFfix_QCD_events_$1_$2_$3_$4 --val-dataset-size 1000 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius $4 --num-blocks $1 -mv-ch $2 -s-ch $3 --spatial-part-only --validation-steps 2000 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_QCD_events_10_16_64_0.8_2025_05_16_19_46_57_48/step_50000_epoch_12.ckpt
20
+
21
+ exit 0
22
+ EOT
23
+
24
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
25
+ # bash jobs/base_training_different_datasets/aug/lgatr_QCD.sh 10 16 64 0.8
jobs/base_training_different_datasets/aug_IRC_S/lgatr_700_07.sh ADDED
@@ -0,0 +1,27 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesSVJTrain # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/7DTrLGATr_out_$1_$2_$3_R$4.log
11
+ #SBATCH --error=jobs/vega/7DTrLGATr_err_$1_$2_$3_R$4.log
12
+
13
+
14
+ source env.sh
15
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
16
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
17
+ nvidia-smi
18
+
19
+ srun singularity exec -B /work -B /pnfs -B /t3home -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train3_PU_PFfix_part0/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part1/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part2/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part3/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part4/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part5/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part6/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part7/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part8/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak -val Delphes_020425_train3_PU_PFfix_part9/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name GP_IRC_S_LGATr_training_NoPID_Delphes_PU_PFfix_700_07_$1_$2_$3_$4 --val-dataset-size 10 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius $4 --num-blocks $1 -mv-ch $2 -s-ch $3 --spatial-part-only --validation-steps 2000 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_700_07_10_16_64_0.8_2025_05_16_19_44_46_795/step_50000_epoch_12.ckpt -irc
20
+
21
+ exit 0
22
+ EOT
23
+
24
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
25
+ # bash jobs/base_training_different_datasets/aug_IRC_S/lgatr_700_07.sh 10 16 64 0.8
26
+
27
+
jobs/base_training_different_datasets/aug_IRC_S/lgatr_700_07_and_900_03.sh ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesSVJTrain # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/700and900_DTrLGATr_out_$1_$2_$3_R$4.log
11
+ #SBATCH --error=jobs/vega/700and900_DTrLGATr_err_$1_$2_$3_R$4.log
12
+
13
+
14
+ source env.sh
15
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
16
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
17
+ nvidia-smi
18
+
19
+ srun singularity exec -B /work -B /pnfs -B /t3home -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train3_PU_PFfix_part0/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part1/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part2/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part3/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part4/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part5/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part6/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part7/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part8/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train3_PU_PFfix_part9/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name GP_IRC_S_LGATr_training_NoPID_Delphes_PU_PFfix_700_07_AND_900_03_$1_$2_$3_$4 --val-dataset-size 10 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius $4 --num-blocks $1 -mv-ch $2 -s-ch $3 --spatial-part-only --validation-steps 2000 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_700_07_AND_900_03_10_16_64_0.8_2025_05_16_21_04_26_991/step_50000_epoch_6.ckpt -irc
20
+
21
+ exit 0
22
+ EOT
23
+
24
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
25
+ # bash jobs/base_training_different_datasets/aug_IRC_S/lgatr_700_07_and_900_03.sh 10 16 64 0.8
26
+
jobs/base_training_different_datasets/aug_IRC_S/lgatr_700_07_and_900_03_and_QCD.sh ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesSVJTrain # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/Train_SB1_DTrLGATr_out_$1_$2_$3_R$4.log
11
+ #SBATCH --error=jobs/vega/Train_SB1_DTrLGATr_err_$1_$2_$3_R$4.log
12
+
13
+
14
+ source env.sh
15
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
16
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
17
+ nvidia-smi
18
+
19
+ srun singularity exec -B /work -B /pnfs -B /t3home -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train3_PU_PFfix_part0/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part1/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part2/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part3/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part4/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part5/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part6/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part7/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part8/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak QCDtrain_part0/qcd_test_0 QCDtrain_part0/qcd_test_1 QCDtrain_part0/qcd_test_2 QCDtrain_part0/qcd_test_3 QCDtrain_part0/qcd_test_4 QCDtrain_part0/qcd_test_5 QCDtrain_part0/qcd_test_6 QCDtrain_part0/qcd_test_7 QCDtrain_part0/qcd_test_8 -val Delphes_020425_train3_PU_PFfix_part9/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name GP_IRC_S_LGATr_training_NoPID_Delphes_PU_PFfix_700_07_AND_900_03_AND_QCD_$1_$2_$3_$4 --val-dataset-size 10 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius $4 --num-blocks $1 -mv-ch $2 -s-ch $3 --spatial-part-only --validation-steps 2000 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_700_07_AND_900_03_AND_QCD_10_16_64_0.8_2025_05_16_21_04_26_937/step_50000_epoch_4.ckpt -irc
20
+
21
+ exit 0
22
+ EOT
23
+
24
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
25
+ # bash jobs/base_training_different_datasets/aug_IRC_S/lgatr_700_07_and_900_03_and_QCD.sh 10 16 64 0.8
jobs/base_training_different_datasets/aug_IRC_S/lgatr_QCD.sh ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesSVJTrain # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/QCDDTrLGATr_out_$1_$2_$3_R$4.log
11
+ #SBATCH --error=jobs/vega/QCDDTrLGATr_err_$1_$2_$3_R$4.log
12
+
13
+
14
+ source env.sh
15
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
16
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
17
+ nvidia-smi
18
+
19
+ srun singularity exec -B /work -B /pnfs -B /t3home -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train QCDtrain_part0/qcd_test_0 QCDtrain_part0/qcd_test_1 QCDtrain_part0/qcd_test_2 QCDtrain_part0/qcd_test_3 QCDtrain_part0/qcd_test_4 QCDtrain_part0/qcd_test_5 QCDtrain_part0/qcd_test_6 QCDtrain_part0/qcd_test_7 QCDtrain_part0/qcd_test_8 -val QCDtrain_part0/qcd_test_9 -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name GP_IRC_S_LGATr_training_NoPID_Delphes_PU_PFfix_QCD_events_$1_$2_$3_$4 --val-dataset-size 10 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius $4 --num-blocks $1 -mv-ch $2 -s-ch $3 --spatial-part-only --validation-steps 2000 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_QCD_events_10_16_64_0.8_2025_05_16_19_46_57_48/step_50000_epoch_12.ckpt -irc
20
+
21
+ exit 0
22
+ EOT
23
+
24
+
25
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
26
+ # bash jobs/base_training_different_datasets/aug_IRC_S/lgatr_QCD.sh 10 16 64 0.8
jobs/base_training_different_datasets/aug_IRC_SN/lgatr_700_07.sh ADDED
@@ -0,0 +1,27 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesSVJTrain # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/7DTrLGATr_out_$1_$2_$3_R$4.log
11
+ #SBATCH --error=jobs/vega/7DTrLGATr_err_$1_$2_$3_R$4.log
12
+
13
+
14
+ source env.sh
15
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
16
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
17
+ nvidia-smi
18
+
19
+ srun singularity exec -B /work -B /pnfs -B /t3home -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train3_PU_PFfix_part0/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part1/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part2/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part3/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part4/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part5/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part6/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part7/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part8/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak -val Delphes_020425_train3_PU_PFfix_part9/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name GP_IRC_SN_LGATr_training_NoPID_Delphes_PU_PFfix_700_07_$1_$2_$3_$4 --val-dataset-size 10 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius $4 --num-blocks $1 -mv-ch $2 -s-ch $3 --spatial-part-only --validation-steps 2000 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_700_07_10_16_64_0.8_2025_05_16_19_44_46_795/step_50000_epoch_12.ckpt -irc
20
+
21
+ exit 0
22
+ EOT
23
+
24
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
25
+ # bash jobs/base_training_different_datasets/aug_IRC_S/lgatr_700_07.sh 10 16 64 0.8
26
+
27
+
jobs/base_training_different_datasets/aug_IRC_SN/lgatr_700_07_and_900_03.sh ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesSVJTrain # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/700and900_DTrLGATr_out_$1_$2_$3_R$4.log
11
+ #SBATCH --error=jobs/vega/700and900_DTrLGATr_err_$1_$2_$3_R$4.log
12
+
13
+
14
+ source env.sh
15
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
16
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
17
+ nvidia-smi
18
+
19
+ srun singularity exec -B /work -B /pnfs -B /t3home -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train3_PU_PFfix_part0/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part1/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part2/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part3/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part4/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part5/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part6/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part7/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part8/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train3_PU_PFfix_part9/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name GP_IRC_SN_LGATr_training_NoPID_Delphes_PU_PFfix_700_07_AND_900_03_$1_$2_$3_$4 --val-dataset-size 10 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius $4 --num-blocks $1 -mv-ch $2 -s-ch $3 --spatial-part-only --validation-steps 2000 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_700_07_AND_900_03_10_16_64_0.8_2025_05_16_21_04_26_991/step_50000_epoch_6.ckpt -irc
20
+
21
+ exit 0
22
+ EOT
23
+
24
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
25
+ # bash jobs/base_training_different_datasets/aug_IRC_S/lgatr_700_07_and_900_03.sh 10 16 64 0.8
26
+
jobs/base_training_different_datasets/aug_IRC_SN/lgatr_700_07_and_900_03_and_QCD.sh ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesSVJTrain # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/Train_SB1_DTrLGATr_out_$1_$2_$3_R$4.log
11
+ #SBATCH --error=jobs/vega/Train_SB1_DTrLGATr_err_$1_$2_$3_R$4.log
12
+
13
+
14
+ source env.sh
15
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
16
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
17
+ nvidia-smi
18
+
19
+ srun singularity exec -B /work -B /pnfs -B /t3home -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train3_PU_PFfix_part0/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part1/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part2/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part3/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part4/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part5/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part6/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part7/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part8/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak QCDtrain_part0/qcd_test_0 QCDtrain_part0/qcd_test_1 QCDtrain_part0/qcd_test_2 QCDtrain_part0/qcd_test_3 QCDtrain_part0/qcd_test_4 QCDtrain_part0/qcd_test_5 QCDtrain_part0/qcd_test_6 QCDtrain_part0/qcd_test_7 QCDtrain_part0/qcd_test_8 -val Delphes_020425_train3_PU_PFfix_part9/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name GP_IRC_SN_LGATr_training_NoPID_Delphes_PU_PFfix_700_07_AND_900_03_AND_QCD_$1_$2_$3_$4 --val-dataset-size 10 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius $4 --num-blocks $1 -mv-ch $2 -s-ch $3 --spatial-part-only --validation-steps 2000 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_700_07_AND_900_03_AND_QCD_10_16_64_0.8_2025_05_16_21_04_26_937/step_50000_epoch_4.ckpt -irc
20
+
21
+ exit 0
22
+ EOT
23
+
24
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
25
+ # bash jobs/base_training_different_datasets/aug_IRC_S/lgatr_700_07_and_900_03_and_QCD.sh 10 16 64 0.8
jobs/base_training_different_datasets/aug_IRC_SN/lgatr_900_03.sh ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesTrainSVJ # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/Dtr_lgatr_aug.out
11
+ #SBATCH --error=jobs/vega/Dtr_lgatr_aug.err
12
+
13
+ source env.sh
14
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
15
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
16
+ nvidia-smi
17
+
18
+ srun singularity exec -B /t3home/gkrzmanc/ -B /work/gkrzmanc/ -B /pnfs/psi.ch/cms/trivcat/store/user/gkrzmanc -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train2_PU_PFfix_part0/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part1/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part2/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part3/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part4/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part5/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part6/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part7/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak Delphes_020425_train2_PU_PFfix_part8/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -val Delphes_020425_train2_PU_PFfix_part9/SVJ_mZprime-900_mDark-20_rinv-0.3_alpha-peak -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name Delphes_Aug_IRCSplit_50k_SN_from3kFT --val-dataset-size 150 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius 0.8 --num-blocks 10 -mv-ch 16 -s-ch 64 --spatial-part-only --validation-steps 60 --no-pid --augment-soft-particles --load-model-weights train/Delphes_Aug_IRCSplit_50k_SN_2025_05_12_13_57_45_477/step_3060_epoch_1.ckpt --num-workers 0 -irc
19
+
20
+
21
+ exit 0
22
+ EOT
23
+
24
+ # bash jobs/IRC_training/start_at_50k/Delphes_training_t3_NoPID_augment_IRCSN.sh
25
+ ####
jobs/base_training_different_datasets/aug_IRC_SN/lgatr_QCD.sh ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesSVJTrain # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/QCDDTrLGATr_out_$1_$2_$3_R$4.log
11
+ #SBATCH --error=jobs/vega/QCDDTrLGATr_err_$1_$2_$3_R$4.log
12
+
13
+
14
+ source env.sh
15
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
16
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
17
+ nvidia-smi
18
+
19
+ srun singularity exec -B /work -B /pnfs -B /t3home -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train QCDtrain_part0/qcd_test_0 QCDtrain_part0/qcd_test_1 QCDtrain_part0/qcd_test_2 QCDtrain_part0/qcd_test_3 QCDtrain_part0/qcd_test_4 QCDtrain_part0/qcd_test_5 QCDtrain_part0/qcd_test_6 QCDtrain_part0/qcd_test_7 QCDtrain_part0/qcd_test_8 -val QCDtrain_part0/qcd_test_9 -net src/models/LGATr/lgatr.py -bs 8 --gpus 0 --run-name GP_IRC_SN_LGATr_training_NoPID_Delphes_PU_PFfix_QCD_events_$1_$2_$3_$4 --val-dataset-size 10 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius $4 --num-blocks $1 -mv-ch $2 -s-ch $3 --spatial-part-only --validation-steps 2000 --no-pid --augment-soft-particles --load-model-weights train/LGATr_training_NoPID_Delphes_PU_PFfix_QCD_events_10_16_64_0.8_2025_05_16_19_46_57_48/step_50000_epoch_12.ckpt -irc
20
+
21
+ exit 0
22
+ EOT
23
+
24
+
25
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
26
+ # bash jobs/base_training_different_datasets/aug_IRC_S/lgatr_QCD.sh 10 16 64 0.8
jobs/base_training_different_datasets/lgatr_700_07.sh ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/bin/bash
2
+ sbatch <<EOT
3
+ #!/bin/bash
4
+ #SBATCH --partition=gpu # Specify the partition
5
+ #SBATCH --mem=25000 # Request 10GB of memory
6
+ #SBATCH --time=48:00:00
7
+ #SBATCH --job-name=DelphesSVJTrain # Name the job
8
+ #SBATCH --gres=gpu:1
9
+ #SBATCH --account=gpu_gres # Specify the account
10
+ #SBATCH --output=jobs/vega/DTrLGATr_out_$1_$2_$3_R$4.log
11
+ #SBATCH --error=jobs/vega/DTrLGATr_err_$1_$2_$3_R$4.log
12
+
13
+
14
+ source env.sh
15
+ export APPTAINER_TMPDIR=/work/gkrzmanc/singularity_tmp
16
+ export APPTAINER_CACHEDIR=/work/gkrzmanc/singularity_cache
17
+ nvidia-smi
18
+
19
+ srun singularity exec -B /work -B /pnfs -B /t3home -H /t3home/gkrzmanc --nv docker://gkrz/lgatr:v3 python -m src.train -train Delphes_020425_train3_PU_PFfix_part0/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part1/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part2/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part3/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part4/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part5/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part6/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part7/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak Delphes_020425_train3_PU_PFfix_part8/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak -val Delphes_020425_train3_PU_PFfix_part9/SVJ_mZprime-700_mDark-20_rinv-0.7_alpha-peak -net src/models/LGATr/lgatr.py -bs 20 --gpus 0 --run-name LGATr_training_NoPID_Delphes_PU_PFfix_700_07_$1_$2_$3_$4 --val-dataset-size 1000 --num-steps 200000 --attr-loss-weight 0.1 --coord-loss-weight 0.1 --beta-type pt+bc --gt-radius $4 --num-blocks $1 -mv-ch $2 -s-ch $3 --spatial-part-only --validation-steps 2000 --no-pid
20
+
21
+ exit 0
22
+ EOT
23
+
24
+ # Args: n_blocks mv_channels s_channels radius (default: 10, 16, 64, 0.8)
25
+ # bash jobs/base_training_different_datasets/lgatr_700_07.sh 10 16 64 0.8
26
+