Neocortex CS-2 uses Slurm for interactive and batch workloads. The SDFlex host nodes support data preparation, validation, compilation, training, and evaluation workflows for the Cerebras CS-2 systems.
Before submitting a job, identify the appropriate PSC project and select its Unix group:
projects groups newgrp GRANT_ID
Common Slurm commands include:
squeue -u $USER # Show your queued and running jobs squeue -j JOB_ID # Show one job scancel JOB_ID # Cancel a job scontrol show job JOB_ID # Display detailed job information
Start an interactive job on either available SDF node with:
interact
To request a specific SDF node, use:
srun --nodelist=sdf-1 --pty bash -i
For CPU-based validation or compilation, request an SDF host node and start the PSC-provided Cerebras container:
srun --pty \ --cpus-per-task=28 \ --kill-on-bad-exit \ singularity shell \ --cleanenv \ --bind /local1/cerebras/data,/local2/cerebras/data,/local3/cerebras/data,/local4/cerebras/data,$PROJECT \ /ocean/neocortex/cerebras/cbcore_latest.sif
A CS-2 training or evaluation job must request a Cerebras accelerator. A minimal batch-script header is:
#!/usr/bin/bash #SBATCH --job-name=cs2-job #SBATCH --account=GRANT_ID #SBATCH --time=04:00:00 #SBATCH --gres=cs:cerebras:1 #SBATCH --ntasks=7 #SBATCH --cpus-per-task=14 #SBATCH --output=%x-%j.out newgrp GRANT_ID # Add the Cerebras container and application commands here.
Submit and monitor the script with:
sbatch job.sbatch squeue -u $USER
The documented default wall time for normal interactive and batch jobs is four hours, and the maximum wall time is 48 hours. Reservations may use a different duration when approved by the Neocortex team.
See the instructions for running interactive and batch CS-2 jobs, the complete Neocortex compilation and batch-script examples, and the Neocortex reservation and wall-time instructions.