• myhloli's avatar
    fix(pdf_parse): handle blocks without lines and enable bf16 on compatible devices · 2145a8b6
    myhloli authored
    Blocks without lines are now correctly indexed even when they contain textual content rendered
    as images. The sorting logic has been updated to accommodate this scenario. Additionally, the
    LayoutLMv3 model initialization has been enhanced to utilize bfloat16 precision on devices that
    support it, offering potential performance benefits on supported hardware.
    2145a8b6
Name
Last commit
Last update
..
dict2md Loading commit data...
filter Loading commit data...
integrations Loading commit data...
layout Loading commit data...
libs Loading commit data...
model Loading commit data...
para Loading commit data...
pipe Loading commit data...
post_proc Loading commit data...
pre_proc Loading commit data...
resources Loading commit data...
rw Loading commit data...
spark Loading commit data...
tools Loading commit data...
v3 Loading commit data...
__init__.py Loading commit data...
pdf_parse_by_ocr.py Loading commit data...
pdf_parse_by_txt.py Loading commit data...
pdf_parse_union_core.py Loading commit data...
pdf_parse_union_core_v2.py Loading commit data...
user_api.py Loading commit data...