marker Releases
115 releases of datalab-to/marker
- v1.2.1Bugfix from last release
Jan 2, 2025
- v1.2.0LLM mode, better OCR heuristics, faster
This release introduces an optional LLM mode to boost output quality for tables, math, and complex pages. It also improves OCR heuristics for automatic error detection and re-OCRing, while significantly speeding up the layout model.
Jan 2, 2025
- v1.1.0Marker Bugfixes and Improvements to `pdftext`
This release focuses on bug fixes and improvements to the `pdftext` dependency, incorporating heuristic-based segmentation and updating related libraries.
Dec 12, 2024
- v1.0.2Bugfixes - python 3.10 compatibility, quotes, images
Minimal release notes: "- Fix issue with python 3.10 - Fix positions of quote characters - Change default image output type to JPEG for speed "
Dec 3, 2024
- v1.0.1Bugfixes and parsing improvements
A patch release addressing parsing issues such as encoding and image rendering, while adding new processors for blockquotes and nested lists.
Dec 3, 2024
- v1.0.0Marker v1!
Marker v1.0 is a complete rewrite from scratch, introducing a new layout model that doubles performance and a modular architecture for easier customization.
Nov 27, 2024
- v0.3.10Performance improvements, API server
Minimal release notes: "- Improve performance by 10-15% - Add a simple API server for local use-cases"
Oct 31, 2024
- v0.3.9Flatten PDF, fix page separators, fix torch/transformers bugs
This release resolves compatibility issues with newer versions of PyTorch and Transformers, improves page separator visualization, and flattens form fields.
Oct 25, 2024
- v0.3.8Fix table bug
Minimal release notes: "- Fix bug that caused conversion to fail when start_page was set and the document had tables"
Oct 24, 2024
- v0.3.7Undo threads
Oct 23, 2024
- v0.3.6Speedups, bug fixes
Minimal release notes: "- Fix some edge case OCR bugs - ~20% end to end speedup by improving layout and text detection"
Oct 23, 2024
- v0.3.5Fix OCR bugs
Minimal release notes: "- Fix bbox issue with OCR and resizing - Fix issue with layout bboxes missing after OCR"
Oct 23, 2024
- v0.3.4Fix misc bugs
Minimal release notes: "- Ensure we don't have 0 area table boxes - Ensure fullymergedblock gets a valid input"
Oct 22, 2024
- v0.3.3Fix layout bugs
Minimal release notes: "- Improve layout, which improves output quality - Fix header level detection bugs"
Oct 22, 2024
- v0.3.2Fix OOM errors
Minimal release notes: "- Add batch size for table rec model to avoid OOM - Enable configuring batch size - Fix error with debugging"
Oct 21, 2024
- v0.3.1Bugfixes, output quality improvement
Minimal release notes: "- Fix MPS bug with torch 2.5 - Fix heading bug with zero line blocks - Improve output quality when visual boxes and te"
Oct 18, 2024
- v0.3.0Better tables, improved output quality, header levels
This release focuses on enhancing table rendering, markdown output quality, and header detection capabilities. It introduces a custom table model and refines block detection logic for better global formatting.
Oct 17, 2024
- v0.2.17OCR and misc improvements; demo app
This update improves OCR functionality and introduces a GUI demo application for testing conversions. It also adds benchmarking tools and progress indicators for equation detection.
Aug 19, 2024
- v0.2.16Significant speedup
A performance-focused release that delivers significant speedups across different hardware backends by adopting more efficient surya models for layout and text detection.
Jul 12, 2024
- v0.2.15Fix transformers bugs
This release adds support for output pagination and enables MPS usage in the convert script. It also resolves several bugs related to RAM settings and file naming.
Jun 30, 2024