pypmltracker 0.1.0

A comprehensive ML experiment tracking library

MLTracker: A Comprehensive ML Experiment Tracking Library

MLTracker is a powerful and flexible experiment tracking library designed for machine learning projects. It helps data scientists and ML engineers efficiently track, visualize, and manage their experiments. Whether you're training models locally or in the cloud, MLTracker provides all the tools you need to streamline your ML workflow and collaborate with your team.

---

Features ✨

• 📊 Experiment Tracking: Log metrics, parameters, and artifacts during model training.
• 🖥️ System Monitoring: Track CPU, memory, disk, and GPU usage during experiments.
• 🔄 Framework Integrations: Native support for PyTorch, TensorFlow, and scikit-learn.
• 📈 Visualization: Interactive dashboard to visualize and compare experiments.
• 💾 Storage Options: Local storage and cloud storage (AWS S3) support.
• 🌐 API: Client-server architecture for team collaboration.

---

---

🔧 Used Libraries & Tools

• 🔗 PyTorch:
• 🔗 TensorFlow:
• 🔗 scikit-learn:
• ☁️ AWS S3:
• 📈 Plotly:
• 📦 MLTracker:

---

Installation ⚙️

Install MLTracker

To install MLTracker, use the following pip command:

pip install mltracker

Optional Features

You can install specific features based on your needs:

• PyTorch integration:

pip install mltracker[pytorch]

• TensorFlow integration:

pip install mltracker[tensorflow]

• scikit-learn integration:

pip install mltracker[sklearn]

• Cloud storage (AWS S3):

pip install mltracker[cloud]

• All features (PyTorch, TensorFlow, scikit-learn, and Cloud):

pip install mltracker[all]

Quick Start 🚀

1. Initialize an Experiment

import mltracker

experiment = mltracker.Experiment(
    project_name="my_project",
    run_name="first_run",
    config={"learning_rate": 0.01, "batch_size": 32}
)
experiment.log({"accuracy": 0.85, "loss": 0.35})
experiment.log_artifact("model", "model.pkl")
experiment.finish()
dashboard = mltracker.Dashboard()
dashboard.start(open_browser=True)

Framework Integrations 🤖

MLTracker integrates with several popular machine learning frameworks. Below are examples of how to use MLTracker with PyTorch, TensorFlow, and scikit-learn.

import torch
import torch.nn as nn
import mltracker

experiment = mltracker.Experiment(project_name="pytorch_example")
tracker = mltracker.PyTorchTracker(experiment, log_gradients=True)
model = nn.Sequential(nn.Linear(10, 5), nn.ReLU(), nn.Linear(5, 1))
tracker.watch(model)

for epoch in range(10):
    loss = train_step(model, data)
    tracker.track_metrics({"loss": loss})
    val_accuracy = validate(model, val_data)
    tracker.on_epoch_end(epoch, model, {"val_accuracy": val_accuracy})
tracker.save_model(model, "final_model")

System Monitoring 🖥️

Track system resources such as CPU, memory, and disk usage during your experiments.

import mltracker
import time

experiment = mltracker.Experiment(project_name="system_monitoring")
monitor = mltracker.SystemMonitor(experiment)
monitor.start()

for i in range(10):
    time.sleep(1)
    experiment.log({"step": i, "value": i * 2})
monitor.stop()
experiment.finish()

Dashboard 📊

Start the MLTracker web dashboard to visualize and compare your experiments.

import mltracker

dashboard = mltracker.Dashboard(
    storage_dir="./mltracker_data",
    host="127.0.0.1",
    port=8000
)
dashboard.start(open_browser=True)

Team Collaboration 🤝

MLTracker supports client-server architecture to facilitate team collaboration.

• Server Start the server to expose MLTracker functionality via a REST API:

import mltracker

server = mltracker.MLTrackerServer(
    storage_dir="./mltracker_data",
    host="0.0.0.0",
    port=5000,
    api_key="your-secret-api-key"
)
server.start()

• Client 📡 Connect to the remote server and interact with it:

import mltracker

client = mltracker.MLTrackerClient(
    base_url="http://server-address:5000",
    api_key="your-secret-api-key"
)
projects = client.list_projects()
print("Projects:", projects)

License 📜

MLTracker is licensed under the MIT License.

Contributing 🤗

Contributions are welcome! If you’d like to contribute to the development of MLTracker, please feel free to submit a pull request.