pySMLE: Simplify Machine Learning Environments

Stop writing boilerplate. Start training.

pySMLE (you can simply call it SMILE) is a lightweight Python framework that automates the "boring stuff" in Machine Learning projects. It handles configuration parsing, logging setup, and experiment tracking so you can focus on the model.

Why pySMLE?

• Auto-Configuration: yaml files are automatically parsed and injected into your entrypoint. No more hardcoded hyperparameters.
• Instant Logging: All print statements and configs are automatically captured to local logs and remote trackers.
• Remote Monitoring: Native integration with Weights & Biases (WandB) to monitor experiments from anywhere.

Installation

pip install smle

Quickstart

Initialize a Project

Run the CLI tool to generate a template and config file:

smle init

Configuration

pySMLE relies on a simple YAML structure to define hyperparameters, paths, logging options, and integrations. You can configure the smle.yaml file with the hyperparameters and options for your project.

The structure of the smle.yaml file is:

# ---------------------------------------
# pySMLE Configuration (Modify Carefully)
# ---------------------------------------

project: project_name

# ---------------------------
# Logging & Tracking
# ---------------------------

logger:
  dir: logger

wandb:
  entity: your_wandb_account

# ---------------------------------------
# Example of User Section
# ---------------------------------------

seed: seed
device: 'cpu'/'cuda'

training:
    epochs: n_epochs
    lr: lr
    weight_decay: wd
    batch: batch_size

testing:
    batch: batch_size

Note. pySMLE expects your Weights and Biases API key to be in the environment variable WANDB_API_KEY. You can put it in the .env file, but ensure .env is in your .gitignore.

Write Your Code

Use the @app.entrypoint decorator. Your configuration variables are automatically passed via args.

from smle import SMLE()

app = SMLE()

@app.entrypoint
def main(args):
    # 'args' contains your pysmle.yaml configurations
    print(f"Training with learning rate: {args['training']['lr']}")

    # Your logic here...

if __name__ == "__main__":
    app.run()

By default, pySMLE will look for a configuration file named smle.yaml in the current directory. If you would like to use a different name, a different location, or have multiple configuration files for different configurations, you can set the config_file property of pySMLE to the path of your file. You must assign the filename before calling run().

app = SMLE()
app.config_file = "my_file.yaml"
...
app.run()

Run It

python main.py

Contributing

Contributions are welcome!

If you’re interested in helping, please feel free to join our discord server or the dedicated discussion page and ping there your availability.

We can then discuss a possible contribution together, answer any questions, and help you get started!

Please, before opening a pull request, consult our CONTRIBUTING.md

Thank you for your support!

Roadmap

High Priority

• Documentation: Write comprehensive documentation and examples.

Planned Features

• ML Templates: Automated creation of standard project structures.
• Model Tools: Utilities for Neural Network creation, training, and testing.
• Notifications: Email notification system for completed training runs.
• Data Tools: Data exploration and visualization helpers.
• Analysis: Result analysis tools (diagrams, confusion matrices, etc.).
• Integrations: Support for TensorBoard and similar tracking tools.
• Testing: Comprehensive unit and integration tests for the framework.