...
- SDF guide and documentation, particularly on using Jupyter notebooks interactively or through web interface.
- Rachel H's documentation on training GNNs with SALT on SDF.
- SDF web browser interface, where you can launch Jupyter jobs on SLURM
- Fork of training-dataset-dumper git repository
- SALT documentation, puma git repo (used for plotting)
- SLAC GitLab group for training related code
Presentations and useful meetings
...
- SDF web interface > My Interactive Sessions > Services > Jupyter (starts a server via SLURM)
- Choose Custom Singularity Containers > enter the following custom commands:
export SINGULARITY_IMAGE_PATH=/sdf/group/atlas/g/XbbXccTrainingData/singularity/vertexing.sif
function jupyter() { singularity exec --nv -B /sdf,/gpfs,/scratch,/lscratch ${SINGULARITY_IMAGE_PATH} jupyter $@; }
Jupyter Instance > slac-ml/SSAI (using the Jupyter instance developed for SSI 2023)
Producing H5 samples
We are using a custom fork of dataset-dumper, developed for producing h5 files for NN training based on FTAG derivations. The custom fork is modified to store the truth jet pT via AntiKt4TruthDressedWZJets container.
...
You can grant read/write access for SDF GPFS data folder directories to ATLAS group members via
Code Block |
---|
chgrp atgroups <username> # To check user groups cd <your_directory> find . -type d|xargs chmod g+r <your_directory>rx # Need to make all subdirectories available to the group |