Releases: VikParuchuri/surya
Releases · VikParuchuri/surya
Performance Improvements
- Don't pass around heatmaps by default
- Use threadpool vs processpool
- Optimize fn for postprocessing
Flatten table form fields, overlap postprocessing
- Overlap postprocessing if batch size >1
- Flatten in form fields with tables
Transformers 4.46, torch 2.5 fixes
- Fixes cudnn backend issue with torch 2.5
- Fixes issue with sdpa and transformers 4.46
- Flatten form fields into pdf by default for table rec
Threads
Threads cause issues on a small % of devices. Although they do give good speedups on most, supporting them seems like a bad idea.
Layout and detection speedup
Corrected version of a recent release with no deadlocks.
- Layout 30% faster
- Text detection 20% faster
Overlaps inference and postprocessing.
Revert thread changes
There were some issues with threading on certain devices. Will re-release after fixing.
Layout and text detection speedup
Overlap postprocessing with inference.
- 20% text detection speedup
- 30% layout speedup
Fix table recognition bug
There was an issue with columns not being detected properly
Fix bug with MPS and PyTorch 2.5
This caused the table rec and OCR models to crash on MPS. Bug is now fixed.
Misc bugfixes
- Fix issue with loading from folders
- Bump pdftext version
- Fix transformers warning