yujunjun/olmocr
1
0
Fork 0
mirror of https://github.com/allenai/olmocr.git synced 2025-10-09 23:32:02 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fixing paddle ocr to run fast

Browse Source
This commit is contained in:
Jake Poznanski 2025-09-19 16:57:05 +00:00
parent 8f75ea062e
commit cedd4a80cf
1 changed files with 17 additions and 16 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

33
olmocr/bench/runners/run_paddlepaddle.py
View File

@ -1,8 +1,4 @@
import json
import os
from typing import Literal
from openai import OpenAI
from threading import Lock
from paddleocr import PPStructureV3
@ -10,24 +6,29 @@ from paddleocr import PPStructureV3
# Run's paddle paddle as in the docs here: https://huggingface.co/PaddlePaddle/PP-OCRv5_server_det
# text_detection_model_name="PP-OCRv5_server_det",
# and using the PP-StructureV3 pipeline to create markdown
paddle_pipeline = None
paddle_pipeline_lock = Lock()
def run_paddlepaddle(
pdf_path: str,
page_num: int = 1,
**kwargs
) -> str:
pipeline = PPStructureV3(
text_detection_model_name="PP-OCRv5_server_det",
use_doc_orientation_classify=False, # Use use_doc_orientation_classify to enable/disable document orientation classification model
use_doc_unwarping=False, # Use use_doc_unwarping to enable/disable document unwarping module
use_textline_orientation=False, # Use use_textline_orientation to enable/disable textline orientation classification model
device="gpu:0", # Use device to specify GPU for model inference
)
output = pipeline.predict(pdf_path)
global paddle_pipeline
with paddle_pipeline_lock:
if paddle_pipeline is None:
paddle_pipeline = PPStructureV3(
text_detection_model_name="PP-OCRv5_server_det",
use_doc_orientation_classify=False, # Use use_doc_orientation_classify to enable/disable document orientation classification model
use_doc_unwarping=False, # Use use_doc_unwarping to enable/disable document unwarping module
use_textline_orientation=False, # Use use_textline_orientation to enable/disable textline orientation classification model
device="gpu:0", # Use device to specify GPU for model inference
)
output = paddle_pipeline.predict(pdf_path)
for cur_page_0_indexed, res in enumerate(output):
if cur_page_0_indexed == page_num - 1:
print(res.markdown)
return res.markdown
return res.markdown["markdown_texts"]
raise ValueError(f"Did not get markdown for page {page_num}")