Hugging Face's logo

Spaces:
vivekreddyr
/
PaddleOCR-VL-Doc-Editor
Running on Zero

App Files Community
PaddleOCR-VL-Doc-Editor

Commit History

  1. Disable experimental SSR: its proxy intermittently 502s on file uploads Running

    vivekreddyr
    Verified signature
    14dbaa2

  2. Fix 'Upload a document first' when a button is clicked while the upload is still loading

    vivekreddyr
    Verified signature
    be85212

  3. Process a page range in one run and export one PDF; unique field names per page; working vertical fields

    vivekreddyr
    Verified signature
    f26161e

  1. Form PDF: white-mask fields (no duplication), preserve vertical text orientation, set NeedAppearances

    vivekreddyr
    Verified signature
    2084545

  2. Editable-form PDF: transparent fields with thin outline (no opaque yellow patches)

    vivekreddyr
    Verified signature
    2dd1411

  3. Add 'Editable form (fillable fields)' PDF export as default; edit fields in any PDF viewer

    vivekreddyr
    Verified signature
    814f113

  1. Export a corrected PDF (searchable overlay or text-only) instead of markdown/json; track zones per page

    vivekreddyr
    Verified signature
    d2a4ec0

  2. Decode only generated tokens (strip prompt echo from recognized text)

    vivekreddyr
    Verified signature
    a98b2b1

  3. Add hidden /parse_sample API endpoint for smoke testing and API/MCP use

    vivekreddyr
    Verified signature
    e839e68

  4. Shim create_causal_mask kwarg (inputs_embeds->input_embeds) so remote code runs on transformers 4.55

    vivekreddyr
    Verified signature
    a3e2288

  5. Quota-friendlier chunking (12/chunk) and lower table/chart token caps

    vivekreddyr
    Verified signature
    ff5d647

  6. analyze: default to page 0 when a doc is loaded (avoid click-race error)

    vivekreddyr
    Verified signature
    2bf6da3

  7. Chunk GPU recognition into short allocations to fit ZeroGPU free-tier duration cap

    vivekreddyr
    Verified signature
    9ea5eb2

  8. Fix VLM (transformers 4.55.0); add dense text-line detection mode + figure OCR + sample

    vivekreddyr
    Verified signature
    2aa6087

  9. Pin gradio 5.50.0 to keep huggingface-hub <1.0 compatible with transformers 4.57

    vivekreddyr
    Verified signature
    87c33c6

  10. Pin transformers 4.57.6 (remote code needs <5); use torch_dtype

    vivekreddyr
    Verified signature
    6133167

  11. ZeroGPU: force cuda placement via env; use transformers 5.x dtype arg

    vivekreddyr
    Verified signature
    616f984

  12. Initial PaddleOCR-VL doc editor: PDF upload, clickable layout zones, editable VLM recognition

    vivekreddyr
    Verified signature
    89e6b1a

  13. initial commit

    vivekreddyr
    Verified signature
    c3d95ed