marker Releases
115 releases of datalab-to/marker
- v0.2.14Pagination, bug fixes
This update addresses compatibility issues introduced by newer versions of the transformers library, specifically regarding donut models and MPS settings.
Jun 17, 2024
- v0.2.13Fix convert.py bug
May 30, 2024
- v0.2.12Specify page range
Minimal release notes: "- Make it more clear MPS can't be used with convert.py - Specify page range in convert with start_page and max_pages"
May 29, 2024
- v0.2.11Python 3.12 compatibility
Ensures Python 3.12 compatibility by removing the Ray dependency, which results in reduced VRAM usage and increased throughput.
May 28, 2024
- v0.2.10OCR speedups
This release focuses on performance improvements, specifically speeding up OCR and text extraction processes by updating internal libraries and enabling new configuration options.
May 28, 2024
- v0.2.9Speed improvements
Minimal release notes: "- Enable parallel text extraction, with worker count settings - Bump surya version to pull in layout/line segmentation "
May 23, 2024
- v0.2.8Faster OCR
Minimal release notes: "- OCR is now ~2.5x faster, due to improvements in surya"
May 18, 2024
- v0.2.7Speed up inference
This update aims to reduce processing time by leveraging updated versions of the Surya OCR library and relaxing version constraints on the transformers library.
May 17, 2024
- v0.2.6Fix memory leak
Minimal release notes: "- Fix a memory leak (fixed in surya, bumped the version). This caused high CPU memory usage on long docs. - Improve lo"
May 16, 2024
- v0.2.5Marker v2
This is a major version rewrite (v2) introducing significant architectural changes, including image extraction, better document formatting, easier installation, and commercial licensing support, though at the cost of doubled processing time.
May 10, 2024
- v0.5.0
Adds support for a new WebSocket transport and simplifies releases by dropping Linux musl builds. It also removes the compilation of TLS features for embedded platforms.
Oct 1, 2023
- v0.4.8
A maintenance release focused on fixing build issues by upgrading various dependencies including libgit2-sys, tokio, openssl, and h2.
May 26, 2023
- v0.0.1
Feb 17, 2023
- v0.4.7
Minimal release notes: "## What's Changed * fix: enable TCP_NODELAY by default by @rapiz1 in https://github.com/rapiz1/rathole/pull/210 **"
Nov 30, 2022
- v0.4.6
Introduces a configurable retry interval for clients and updates the vergen crate to support SOURCE_DATE_EPOCH.
Nov 29, 2022