Bayes_HEP Instructions

This guide provides step-by-step instructions for running the Bayes_HEP project using Docker or Apptainer (formerly Singularity), both interactively and in batch mode on HPC systems.

---

🚀 Docker Workflow

1. Setup

mkdir Bayes_HEP
cd Bayes_HEP

docker pull cbaillar/bayes_hep:latest

docker run -it --rm -v "$PWD":/workdir -e WORKDIR=/workdir cbaillar/bayes_hep

cp -r /usr/local/share/Bayes_HEP/Examples/New_Project .

---

2. Input Configuration

Edit the following files in Docker_New_Project/input/Rivet/:

• `analyses_list.txt` – List of Rivet analyses to run
  Example:

  pp_7000:
  ATLAS_2010_I882098 d03-x01-y01 d07-x01-y01
  ATLAS_2012_I1188891 d02-x01-y01
  
  pp_13000:
  CMS_2018_I1663452 d03-x01-y01
  

• `parameter_prior_list.dat` – Set parameter prior ranges for sampling/tuning

• `parameter.cmnd` – Default Pythia parameter card used as a template for all runs

---

3. Interactive Execution (Inside Container)

python New_Project/drivers/Rivet_Main.py
python New_Project/drivers/Bayes_Main.py

---

4. Batch Job Submission (Docker)

Before submitting jobs, update generate_design_points.sh in Docker_New_Project/Batch_Rivet/ to match your setup.

Key parameters:

COLLISIONS="pp_7000"        # Space-separated systems, e.g., "pp_7000 pp_13000"
TOTAL_POINTS=5              # Total number of design points
TOTAL_EVENTS=10             # Total number of events
NEVENTS=10                  # Events per job
CPU=5                       # Number of CPUs to use
TARGET_EVENTS_PER_JOB=1000000

Submit jobs:

bash New_Project/Batch_Rivet/Docker/generate_design_points.sh

---

🧠 HPC Workflow (Apptainer/Singularity)

1. Setup

mkdir Bayes_HEP
cd Bayes_HEP

apptainer build bayes_hep.sif docker://cbaillar/bayes_hep:latest

apptainer shell --bind "$PWD":/workdir bayes_hep.sif

cp -r /usr/local/share/Bayes_HEP/Examples/New_Project .

---

2. Input Configuration

Edit the following files in HPC_New_Project/input/Rivet/:

• `analyses_list.txt` – List of Rivet analyses (see example above)
• `parameter_prior_list.dat` – Parameter prior ranges
• `parameter.cmnd` – Default Pythia template

---

3. Interactive Execution (Inside Container)

python New_Project/drivers/Rivet_Main.py
python New_Project/drivers/Bayes_Main.py

---

4. Batch Job Submission (SLURM)

Before submitting jobs, update:

• generate_design_points.sh and other scripts in HPC_New_Project/Batch_Rivet/
• Paths like USER_DIR and any project-specific configurations

Key parameters in generate_design_points.sh:

USER_DIR="/lustre/isaac24/proj/UTK0244/cbaillar"
COLLISIONS="pp_7000 pp_13000"
TOTAL_POINTS=10
TOTAL_EVENTS=100
NEVENTS=50

Submit jobs:

bash New_Project/Batch_Rivet/HPC/generate_design_points.sh

For MCMC calibration (run_bayes.slurm):

USER_DIR="/lustre/isaac24/proj/UTK0244/cbaillar"
COLLISIONS="pp_7000"
MODEL="pythia8"
N_WALKERS=50
NPOOL=10
SAMPLES=10000
RESULT_SIZE=100

Submit:

sbatch New_Project/Batch_Rivet/HPC/run_bayes.slurm

---

📊 SLURM Job Management

• Check job status:

  squeue -u <your_username>

• Cancel all jobs:

  scancel -u <your_username>

---

✅ Tips

• Double-check all .slurm and .sh files for correct paths and parameters.
• SLURM output and error logs will be saved in the locations specified in your scripts.
• Rebuild or restart containers after modifying input files or source code.

---

❓ Need Help?

If you encounter errors:

• Check SLURM logs for job status and error messages
• Ensure all paths and inputs are correct
• Contact your system administrator or project maintainer for support