EasySubmit

A Python library for simplified job scheduling and management on SLURM clusters.

Overview

EasySubmit is a Python-based job scheduling and management system designed to seamlessly integrate with SLURM, a popular cluster management and job scheduling platform. This project aims to simplify the process of submitting, monitoring, and managing jobs on cluster environments through an intuitive Python API.

Key Features

• Simple API: Easy-to-use Python interface for SLURM job submission
• Task Management: Define and configure tasks with type-safe configuration classes
• Batch Scheduling: Submit multiple experiments or jobs with different parameters
• Profiling Support: Optional integration with Scalene for performance profiling
• Flexible Configuration: Comprehensive SLURM configuration options
• Type Safety: Built with modern Python type hints for better development experience

Installation

From PyPI (Recommended)

pip install easysubmit

From Source

# Clone the repository
git clone https://github.com/ysenarath/easysubmit.git
cd easysubmit

# Install in development mode
pip install -e .

Optional Dependencies

For profiling support:

pip install easysubmit[scalene]

Quick Start

Here's a simple example of how to use EasySubmit:

from easysubmit import SLURMCluster, SLURMConfig, Task, TaskConfig
from easysubmit.base import schedule

# Define your task configuration
class ExperimentConfig(TaskConfig):
    name: str = "MyExperiment"
    learning_rate: float = 0.001
    batch_size: int = 32

# Define your task
class Experiment(Task):
    config: ExperimentConfig
    
    def run(self):
        print(f"Running experiment with lr={self.config.learning_rate}")
        # Your experiment code here

# Configure SLURM settings
config = SLURMConfig(
    partition="gpu",
    nodes=1,
    ntasks_per_node=1,
    gres="gpu:1",
    mem="16G"
)

# Create cluster and schedule jobs
cluster = SLURMCluster(config)
experiments = [
    {"name": "MyExperiment", "learning_rate": 0.001, "batch_size": 32},
    {"name": "MyExperiment", "learning_rate": 0.01, "batch_size": 64},
]

schedule(cluster, experiments)

Core Components

Task and TaskConfig

• TaskConfig: Define configuration parameters for your tasks with type safety
• Task: Base class for implementing your computational tasks

SLURM Integration

• SLURMCluster: Interface to SLURM cluster management
• SLURMConfig: Comprehensive SLURM job configuration options

Job Management

• Job: Represents individual jobs in the cluster
• AutoTask: Advanced task automation features

Prerequisites

• Python 3.9 or higher
• Access to a SLURM cluster environment
• SLURM commands (sbatch, squeue, etc.) available in PATH

Examples

Check out the examples/ directory for more comprehensive usage examples:

• examples/slurm_scheduler.py: Basic SLURM job scheduling
• examples/slurm_scheduler_with_profile.py: Job scheduling with profiling
• examples/tasks.py: Task definition examples

License

This project is licensed under the MIT License - see the LICENSE file for details.

Support

• Issues: GitHub Issues
• Documentation: GitHub README
• Source Code: GitHub Repository