1
Fork 0
This repository has been archived on 2026-08-15. You can view files and clone it, but you cannot make any changes to it's state, such as pushing and creating new issues, pull requests or comments.
2026SEL3-project-Brittle_St.../scripts/hpc/train.pbs
Tibo De Peuter 9dce017525 refactor(hpc): migrate to PBS/vsc-venv, reorganise scripts under scripts/hpc/
- Replace Slurm headers with PBS (#PBS) directives
- Replace manual module+venv logic with vsc-venv --activate
- Cluster selection via 'module swap cluster/<name>' before qsub
- MUJOCO_GL=egl retained: required for headless physics sim and video rendering
- Delete deprecated scripts/hpc_install.sh and scripts/hpc_train.slurm
2026-04-04 18:11:38 +02:00

38 lines
1.3 KiB
Bash

#!/bin/bash
# scripts/hpc/train.pbs — Submit PPO training as a batch job with qsub.
#
# Usage (from project root):
# qsub scripts/hpc/train.pbs
#
# To target a specific GPU cluster (default: joltik):
# module swap cluster/accelgor && qsub scripts/hpc/train.pbs
#
# Available GPU clusters: joltik, accelgor, litleo
# Debug / interactive: doduo (CPU) or donphan (interactive GPU)
# ---------------------------------------------------------------------------
# PBS job directives
# ---------------------------------------------------------------------------
#PBS -N brittlestar-ppo
#PBS -l nodes=1:ppn=8:gpus=1
#PBS -l walltime=24:00:00
#PBS -l mem=32gb
#PBS -o runs/pbs_${PBS_JOBID}.out
#PBS -e runs/pbs_${PBS_JOBID}.err
# ---------------------------------------------------------------------------
set -euo pipefail
cd "$PBS_O_WORKDIR"
# Activate the project virtual environment via the official vsc-venv wrapper.
# This loads the modules listed in env/hpc/modules.txt and the per-cluster venv.
module load vsc-venv
source vsc-venv --activate --modules env/hpc/modules.txt
# EGL is required for headless MuJoCo operation — both for the physics
# simulation loop and for rendering videos to disk. Without this, MuJoCo
# attempts to open an X11 display and fails on compute nodes.
export MUJOCO_GL=egl
python src/train.py --config configs/production_training.yaml