surya Releases
166 releases of datalab-to/surya
- v0.10.1Pin pytorch
Pins the PyTorch version to 2.6.0 to address compatibility issues with MPS (Mac) devices.
Jan 30, 2025
- v0.10.0Add LaTeX OCR model
Introduces a new LaTeX OCR model and a Streamlit app for interactive equation recognition.
Jan 29, 2025
- v0.9.3Fix cli script issue
Jan 24, 2025
- v0.9.2Better polygon type checking
Improves how polygons are type checked within the schema to ensure better data integrity.
Jan 24, 2025
- v0.9.1Fix rowspan bug
Addresses a specific bug where row spans were incorrectly excluded from table model predictions.
Jan 23, 2025
- v0.9.0Refactor surya; new table recognition model
A major refactor of the library introducing a new table recognition model and a significantly simplified API.
Jan 22, 2025
- v0.8.3Pin pypdfium2
Pins the `pypdfium2` dependency version to prevent compatibility issues with newer releases.
Jan 1, 2025
- v0.8.2New layout model
Significant performance update to the layout model, making it faster and more accurate.
Dec 30, 2024
- v0.8.1Add bad OCR detection model
Introduces new models for quality control and adds flexibility to layout predictions.
Dec 19, 2024
- v0.8.0Surya Bugfixes and Improvements to `pdftext`
Updates internal dependencies and introduces a new layout model with improved compilation support.
Dec 12, 2024
- v0.7.0New layout model
Performance optimizations designed to reduce overhead and improve processing speed.
Nov 27, 2024
- v0.6.13Performance Improvements
Optimizes internal processing to improve overall performance and reduce resource usage.
Oct 30, 2024
- v0.6.12Flatten table form fields, overlap postprocessing
Improves table handling by flattening form fields and optimizing overlap postprocessing.
Oct 28, 2024
- v0.6.11Transformers 4.46, torch 2.5 fixes
Addresses compatibility issues with PyTorch 2.5 and Transformers 4.46 while improving table recognition defaults.
Oct 25, 2024
- v0.6.10Threads
Removes threading support due to compatibility issues on specific hardware.
Oct 23, 2024
- v0.6.9Layout and detection speedup
Fixes deadlocks in the previous layout release and provides significant performance improvements by overlapping inference and postprocessing.
Oct 23, 2024
- v0.6.8Revert thread changes
Reverts threading changes introduced in the previous release due to device compatibility issues.
Oct 22, 2024
- v0.6.7Layout and text detection speedup
Resolves compatibility issues causing crashes on Apple Silicon (MPS) with PyTorch 2.5.
Oct 22, 2024
- v0.6.6Fix table recognition bug
Fixes a bug preventing proper column detection in table recognition.
Oct 18, 2024
- v0.6.5Fix bug with MPS and PyTorch 2.5
Introduces overlapping of postprocessing with inference to achieve significant speed increases.
Oct 18, 2024