Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

feat: add slurm launcher #133

Merged
merged 1 commit into from
Feb 10, 2025
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 20 additions & 0 deletions stoix/configs/launcher/slurm.yaml
Original file line number Diff line number Diff line change
@@ -0,0 +1,20 @@
# config.yaml
slurm:
account: ~
qos: ~
nodes: 1
gpus_per_node: 1
cpus_per_task: 24
partition: ~
time: 00:05:00
folder: "./slurm_logs"


experiment:
algorithm_exec_files:
- 'stoix/systems/ppo/anakin/ff_ppo.py'
- 'stoix/systems/ppo/anakin/rec_ppo.py'
environments: [gymnax/cartpole, gymnax/mountain_car]
seeds: [42]

experiment_group: example_group_run
83 changes: 83 additions & 0 deletions stoix/slurm_launcher.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,83 @@
import itertools
import os
import subprocess

import hydra
import submitit
from omegaconf import DictConfig


def run_experiment(algorithm_exec_file: str, environment: str, seed: int) -> None:
"""
Runs a single Stoix experiment via a subprocess run.

Args:
algorithm_exec_file: Algorithm/system (e.g. 'dqn', 'ppo') exec file.
e.g. 'stoix/systems/ppo/anakin/ff_ppo.py'
environment: Environment config (e.g. 'gymnax/cartpole or brax/ant')
seed: Random seed for reproducibility
"""

cmd = f"python {algorithm_exec_file} env={environment} arch.seed={seed}"

print(f"Running command: {cmd}")
subprocess.run(cmd, shell=True, check=True)


def filter_none_values(d: dict) -> dict:
"""
Returns a new dictionary containing only the items from the input dictionary
where the value is not None.

Args:
d: The input dictionary.
Returns:
A dictionary with keys whose values are not None.
"""
return {key: value for key, value in d.items() if value is not None}


@hydra.main(version_base="1.2", config_path="./configs/launcher", config_name="slurm")
def main(cfg: DictConfig) -> None:
"""
Main entry point for launching multiple Stoix experiments on SLURM-based cluster.

Args:
cfg: The Hydra-populated configuration object.
"""
# Create the submitit executor for SLURM.
executor = submitit.AutoExecutor(folder=cfg.slurm.folder)

# Build SLURM parameter dictionary. Only pass parameters that are non-None.
# If you pass None to some fields, submitit may ignore them or raise an error,
# so we filter them out where appropriate.
slurm_params = {
"nodes": cfg.slurm.nodes,
"gpus_per_node": cfg.slurm.gpus_per_node,
"cpus_per_task": cfg.slurm.cpus_per_task,
"time": cfg.slurm.time,
"chdir": os.getcwd(),
"slurm_account": cfg.slurm.account,
"slurm_qos": cfg.slurm.qos,
"slurm_partition": cfg.slurm.partition,
}
slurm_params = filter_none_values(slurm_params)

# Update the executor with SLURM parameters
executor.update_parameters(
slurm_job_name=cfg.experiment_group, slurm_additional_parameters=slurm_params
)

# Prepare the Cartesian product of algorithm_execs, environments, seeds.
jobs = []
with executor.batch():
for algorithm_exec, env, seed in itertools.product(
cfg.experiment.algorithm_exec_files, cfg.experiment.environments, cfg.experiment.seeds
):
print(f"Submitting job for {algorithm_exec.split('/')[-1]} on {env} with seed {seed}.")
job = executor.submit(run_experiment, algorithm_exec, env, seed)
jobs.append(job)


if __name__ == "__main__":
main()