llama-mlx-pipeline

Llama MLX Pipeline (llama-mlx-pipeline) provides tools and components for building efficient data processing pipelines, specifically optimized for Apple Silicon using the MLX framework. It focuses on tasks like data scraping, chunking, and feature extraction.

Key Features

• MLX Optimization: Designed to leverage Apple's MLX framework for high performance on M-series chips.
• Data Scraping: Includes components for fetching data from various sources (scraper.py).
• Data Chunking: Provides utilities for splitting data into manageable chunks (chunker.py).
• Feature Extraction: Contains tools for extracting relevant features or information (extractor.py).
• Pipeline Core: A central module (core.py) likely orchestrates pipeline stages.
• Configurable: Supports configuration via config.py.

Installation

pip install llama-mlx-pipeline
# Or install directly from GitHub for the latest version:
# pip install git+https://github.com/llamasearchai/llama-mlx-pipeline.git

Usage

(Usage examples demonstrating pipeline creation and execution will be added here.)

# Placeholder for Python client usage
# from llama_mlx_pipeline import PipelineBuilder, MLXConfig

# config = MLXConfig.load("path/to/config.yaml")
# builder = PipelineBuilder(config)

# pipeline = builder.add_scraper(source="web", url="...") \
#                   .add_chunker(size=512) \
#                   .add_extractor(model="bert-base") \
#                   .build()

# results = pipeline.run(input_data="...")
# print(results)

Architecture Overview

graph TD
    A[Input Data Source] --> B{Scraper (scraper.py)};
    B --> C{Chunker (chunker.py)};
    C --> D{Extractor (extractor.py)};
    D --> E[Processed Output];

    F[Pipeline Orchestrator (core.py)] -- Manages --> B;
    F -- Manages --> C;
    F -- Manages --> D;

    G[Configuration (config.py)] -- Configures --> F;
    G -- Configures --> B;
    G -- Configures --> C;
    G -- Configures --> D;

    subgraph MLX Optimized Components
        direction LR
        B;
        C;
        D;
    end

    style F fill:#f9f,stroke:#333,stroke-width:2px

1. Input: Data enters the pipeline.
2. Scraper: Fetches or loads the initial data.
3. Chunker: Splits data into smaller pieces.
4. Extractor: Processes chunks to extract features or information.
5. Output: The final processed data is produced.
6. Orchestrator: The core.py module likely manages the flow and execution of these stages, configured by config.py.

Configuration

(Details on configuring pipeline stages, MLX settings, etc., will be added here.)

Development

Setup

# Clone the repository
git clone https://github.com/llamasearchai/llama-mlx-pipeline.git
cd llama-mlx-pipeline

# Install in editable mode with development dependencies
pip install -e ".[dev]"

Testing

pytest tests/

Contributing

Contributions are welcome! Please refer to CONTRIBUTING.md and submit a Pull Request.

License

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