surya Releases
166 releases of datalab-to/surya
- v3.0.5v3.0.5 | Introduce fine-grained restrictions for tags and correspondents
This update introduces fine-grained control over document tagging and correspondence, alongside new AI capabilities for Azure OpenAI and external data fetching.
Jun 16, 2025
- v0.14.6Fix invalid math tags
Improved output formatting by cleaning up math tags to remove unwanted HTML tags and formatting artifacts.
Jun 11, 2025
- v0.14.5Small fix
Jun 2, 2025
- v0.14.4Misc fixes
This release includes performance optimizations and bug fixes, allowing users to control text repetition and adjust chunk sizes for better hardware compatibility.
Jun 2, 2025
- v0.14.3Improve performance on CPU, MPS, older GPUs
Optimized the library for older hardware and CPUs by fixing attention mechanisms and adding input truncation for predictable memory usage.
May 29, 2025
- v3.1.3
May 28, 2025
- v0.14.2Fix bug with large images
Resolved a critical bug affecting large images on specific hardware like MPS, along with general logging improvements.
May 20, 2025
- v0.14.1OCR accuracy improvements
Updated the OCR model to be more robust and accurate, reducing errors in processing.
May 16, 2025
- v0.14.0Surya OCR 3
Major architectural overhaul introducing Surya OCR 3, trained on significantly more data. This release focuses on performance, expanding language support, and improving handling of complex layouts like inline math.
May 13, 2025
- v3.1.2
Apr 15, 2025
- v0.13.1Bugfixes
Bugfix release addressing issues with inline prediction and image mapping.
Mar 26, 2025
- r1.5.0PaddleSpeech r1.5.0
PaddleSpeech r1.5.0 is a major update focusing on the upgrade from Paddle 2.5.1 to Paddle 3.0.0-beta. It includes new features like AudioTools toolkit support and DAC loss functions, hardware support for GCU/NPU/MLU, and extensive adaptation work to ensure compatibility.
Mar 5, 2025
- v0.13.0New text detection model
Introduces a new text detection model designed to detect more text across various PDF documents to improve overall OCR performance.
Feb 28, 2025
- v0.12.1Improved inline math model
Improves the inline math recognition capabilities by deploying a new model.
Feb 24, 2025
- v1.3Open-Sora V1.3 Release
Open-Sora V1.3 introduces improvements in model architecture and generation quality, continuing to support image-to-video generation and video extension with expanded resolution and frame length support.
Feb 21, 2025
- v0.12.0Model downloads with S3
Enhances model loading reliability and removes GUI dependencies by implementing S3 support and switching to headless OpenCV.
Feb 19, 2025
- v0.11.1Fix inline detection bug
Feb 13, 2025
- v0.11.0
Adds inline math detection, benchmarks against Textract/Google Cloud Vision, and introduces XLA support for TPUs.
Feb 10, 2025
- v0.10.3Fix height issue
Fixes a critical height issue in text detection that caused bounding boxes to be misaligned in tall images.
Feb 6, 2025
- v0.10.2Fix pytorch 2.6 bug
Resolves compatibility issues with PyTorch 2.6 on MPS (Mac) devices.
Jan 31, 2025