wyolo 1.0.0

Professional YOLO training library with MLOps integration

wyolo: Professional YOLO MLOps Framework

wyolo is a high-performance, enterprise-grade Python library designed to orchestrate the entire lifecycle of YOLO and RT-DETR models. By integrating native MLOps capabilities, wyolo transforms experimental computer vision into production-ready automated pipelines.

---

1. 🚶 Diagram Walkthrough

The following diagram illustrates the high-level execution flow from user request to model registration.

graph TD
    A[User/Worker Request] --> B{Configuration Loader}
    B --> C[Environment Validator]
    C -->|Success| D[GPU/Batch Optimizer]
    C -->|Failure| E[Error Capture & Notification]
    D --> F[Dataset Synchronization - DVC]
    F --> G[YOLO/RT-DETR Training Loop]
    G --> H[Metrics & Logs - MLflow]
    H --> I[Artifact Storage - MinIO/S3]
    I --> J[Model Registry]
    J --> K[Execution Completed]

---

2. 🗺️ System Workflow

Detailed sequence of events during a critical training operation, highlighting the interaction between core components.

sequenceDiagram
    participant U as User/Worker
    participant TW as TrainerWrapper
    participant MM as MLflowManager
    participant ULY as Ultralytics Engine
    participant S3 as MinIO Storage

    U->>TW: Initialize(config)
    TW->>MM: Setup Experiment & Run
    TW->>ULY: Start Training(hyperparameters)
    loop Every Epoch
        ULY-->>TW: Metrics Update
        TW->>MM: Log Metrics (mAP, Loss)
    end
    ULY-->>TW: Final Model (.pt)
    TW->>S3: Upload Artifacts
    TW->>MM: Register Model Version
    MM-->>U: Final Status & Registry Link

---

3. 🏗️ Architecture Components

Static structure and dependencies of the wyolo ecosystem.

mindmap
  root((wyolo System))
    Core Layer
      TrainerWrapper
      MLflowManager
      GPUUtils
    App Layer (Worker)
      State Machine
      Check States (GPU, Dataset)
      Train State
    External Integrations
      Ultralytics (YOLOv8, RT-DETR)
      MLflow (Tracking, Registry)
      DVC (Data Versioning)
      Redis (Messaging)
    Infrastructure
      Docker (Containerization)
      MinIO/S3 (Storage)

---

4. ⚙️ Container Lifecycle

a. Build Process

The construction of the production-ready artifact follows a multi-stage approach:

1. Base Image Selection: Utilizes NVIDIA CUDA-enabled Python images for GPU acceleration.
2. Dependency Resolution: Installation of system-level libraries (libgl1-mesa-glx) and project dependencies via pip.
3. Code Injection: Surgical copying of the src/ directory and configuration templates.
4. Environment Sanitization: Setting of non-root users and file permissions for secure execution.

b. Runtime Process

From instantiation to operational readiness:

1. Entrypoint Execution: The container starts via train_service.sh.
2. State Initialization: The main.py state machine loads the specific configuration.
3. Resource Discovery: Automatic detection of CUDA devices and available VRAM.
4. Data Mounting: CIFS/NFS mounts or DVC pulls to ensure dataset availability.
5. Steady State: The worker enters the training loop, reporting heartbeats to Redis/MLflow.

---

5. 📂 File-by-File Guide

Path	Purpose
src/wyolo/core/	Core logic for trainer wrappers, MLflow management, and GPU utilities.
src/wyolo/app/	Production worker implementation using a robust state-machine architecture.
src/wyolo/trainer/	Specialized model-type implementations and elemental DTOs.
src/wyolo/docker/	Orchestration scripts and requirements for containerized environments.
pyproject.toml	Project metadata, build system configuration, and dependency locks.
Makefile	Standardized automation for installation, testing, and documentation.
tests/	Comprehensive test suite ensuring behavioral and structural integrity.

---

🛠 Technical Stack

• Engine: Ultralytics (YOLOv8, RT-DETR).
• Tracking: MLflow.
• Versioning: DVC.
• Backend: Python 3.8+ with Loguru for logging.
• Containerization: Docker with NVIDIA Container Toolkit.
• Quality: Pytest, Pylint, Bandit.

---

🚀 Installation & Setup

Environment Preparation

# Clone the repository
git clone https://github.com/wisrovi/wyoloservice2_worker.git
cd wyoloservice2_worker

# Create and activate virtual environment
python -m venv venv
source venv/bin/activate

Automation with Makefile

# Install all dependencies (including dev)
make install

# Run security and linting checks
make lint

# Execute test suite
make test

---

📈 Configuration & Usage

Configuration (YAML)

Create a config.yaml to define your experiment:

model: yolov8n.pt
type: detect
train:
  epochs: 100
  imgsz: 640
mlflow:
  uri: "http://your-mlflow-server"

Execution

# Run training via CLI
wyolo-train --config_path config.yaml --fitness map50

---

👨‍💻 Author

William Rodríguez - wisrovi
Technology Evangelist & AI Solutions Architect
LinkedIn Profile

---

📄 Bibliography & Source Resources

• Ultralytics Documentation
• MLflow Guide
• DVC Official Tutorials