Lift is a vision-language document parsing model developed by DataLab, designed for document understanding, structured document reconstruction, layout analysis, and intelligent content extraction. This repository...
Fuente del modelo
Descripción de la fuente
Lift is a vision-language document parsing model developed by DataLab, designed for document understanding, structured document reconstruction, layout analysis, and intelligent content extraction. This repository contains GGUF quantized variants of the model optimized for efficient local inference using llama.cpp.
Unlike traditional OCR models that primarily focus on recognizing text, Lift emphasizes reconstructing complete document structure while preserving reading order, hierarchy, tables, figures, mathematical expressions, forms, and semantic relationships between document elements. The model enables complex documents to be transformed into structured machine-readable representations suitable for downstream AI workflows.
The quantized formats significantly reduce memory requirements while maintaining strong document parsing capability, making the model suitable for enterprise document intelligence, retrieval systems, and large-scale document conversion pipelines.
Fuentes
1 fuenteVerificado 6 ago
Artefactos del modelo
4 artefactosExtractos de fuentes
2 extractosThis repository provides various GGUF quantized versions of the Lift model optimized for efficient local inference using llama.cpp.
Lift is trained with an emphasis on document parsing, layout understanding, semantic reconstruction, and multimodal document intelligence across diverse document collections.
Document Parsing : Reconstructs complete document structure while preserving semantic organization.
Layout Understanding : Identifies relationships between headings, paragraphs, tables, figures, and other document elements.
Structured Document Conversion : Converts complex documents into machine-readable structured representations.
Semantic Content Extraction : Extracts contextual information while maintaining document hierarchy.
Complex Layout Processing : Handles technical documents, reports, books, forms, and visually rich page layouts.
Efficient Local Deployment : Quantized variants enable practical document intelligence workloads on consumer hardware.
./llama-mtmd-cli \
-m SandlogicTechnologies/Lift_IQ4_NL.gguf \
--mmproj SandlogicTechnologies/lift.mmproj-f16.gguf \
--image technical_report.pdf \
-p "Convert this document into structured Markdown while preserving headings, tables, and figures."
Document Conversion : Transform complex documents into structured machine-readable formats.
Layout-Aware Processing : Preserve document hierarchy, formatting, and reading order during extraction.
Knowledge Base Generation : Prepare structured documents for enterprise search and retrieval systems.
Technical Document Processing : Parse scientific papers, manuals, books, reports, and technical documentation.
Enterprise Document Intelligence : Build automated document understanding and indexing workflows.
RAG Data Preparation : Generate structured representations suitable for Retrieval-Augmented Generation systems.
These quantized models are based on the original work by the DataLab development team.
Special thanks to:
llama.cpp open-source community for enabling efficient quantization and inference via the GGUF format.For questions, feedback, or support, please reach out at support@sandlogic.com or visit https://www.sandlogic.com/
lift-IQ4_XS.gguf
gguf · 4,99 GB · SHA-256 043758a07f12…6dc0 · Hugging Face
--- license: apache-2.0 language: - multilingual base_model: - datalab-to/lift tags: - multimodal - document-understanding - document-parsing - layout-analysis - vision-language-model - document-conversion - image-to-text - efficient-model --- # Lift Lift is a vision-language document parsing model developed by DataLab, designed for document understanding, structured document reconstruction, layout analysis, and intelligent content extraction. This repository contains GGUF quantized variants of the model optimized for efficient local inference using llama.cpp. Unlike traditional OCR models that primarily focus on recognizing text, Lift emphasizes reconstructing complete document structure while preserving reading order, hierarchy, tables, figures, mathematical expressions, forms, and semantic relationships between document elements. The model enables complex documents to be transformed into structured machine-readable representations suitable for downstream AI workflows. The quantized formats significantly reduce memory requirements while maintaining strong document parsing capability, making the model suitable for enterprise document intelligence, retrieval systems, and large-scale document conversion pipelines. --- # Model Overview - **Model Name**: Lift - **Base Model**: datalab-to/lift - **Architecture**: Vision-Language Document Parsing Model - **Parameter Count**: Approximately 8 Billion Parameters - **Modalities**: Text, Image - **Primary Languages**: Multilingual - **Developer**: DataLab - **License**: Apache 2.0 --- # Quantization Formats This repository provides various GGUF quantized versions of the Lift model optimized for efficient local inference using llama.cpp. ## IQ3_M - Size reduction of approx 77.12% (4.21 GB) compared to 16-bit (18.40 GB) - Aggressive 3-bit quantization optimized for memory-efficient document parsing workloads - Suitable for large-scale document conversion pipelines operating on resource-constrained hardware - Enables efficient processing of books, reports, forms, technical papers, and visually rich documents - Preservation of fine document hierarchy, layout relationships, and complex structural elements may reduce compared to higher-precision variants ## IQ4_NL - Size reduction of approx 71.74% (5.20 GB) compared to 16-bit (18.40 GB) - Advanced 4-bit non-linear quanti...
Source context: 274 downloads · 0 likes · Pipeline image-to-text · Repo SandLogicTechnologies/lift-GGUF