SmolDocling-256M-preview-mlx-bf16
| Property | Value |
|---|---|
| Model Size | 256M parameters |
| Framework | MLX |
| Format | BF16 |
| Original Author | ds4sd |
| Model URL | Hugging Face |
What is SmolDocling-256M-preview-mlx-bf16?
SmolDocling-256M-preview-mlx-bf16 is a specialized document understanding model optimized for Apple's MLX framework. It's designed to convert document images into structured formats using the Docling framework, particularly excelling at handling tables and lists within documents.
Implementation Details
The model is implemented using MLX-VLM version 0.1.18 and requires Python 3.12 or higher. It utilizes the docling-core framework for document processing and supports both local and URL-based image inputs.
- Optimized for BF16 precision
- Supports streaming generation
- Integrated with docling-core for document structure parsing
- Compatible with PIL for image processing
Core Capabilities
- Document structure recognition
- Table and list extraction
- HTML and Markdown export
- Embedded image handling
- Real-time token generation
Frequently Asked Questions
Q: What makes this model unique?
This model stands out for its optimization for Apple's MLX framework and its ability to process document images into structured formats while maintaining a relatively small parameter count of 256M.
Q: What are the recommended use cases?
The model is ideal for document parsing tasks, particularly when working with documents containing tables and lists. It's especially useful in environments where MLX optimization is beneficial, such as Apple Silicon hardware.
