opensource/MinerU
3
0
Fork 0
mirror of https://github.com/opendatalab/MinerU.git synced 2026-10-29 19:07:28 +08:00
Code Issues Packages Projects Releases Wiki Activity

feat: add client-side output generation option for middle JSON finalization

Browse Source
This commit is contained in:
myhloli
2026-05-27 20:47:04 +08:00
parent f51547c9d2
commit baa7274a0f
1 changed files with 36 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files
mineru/backend/hybrid/hybrid_analyze.py
+36 -13
View File
@@ -13,6 +13,7 @@ from mineru_vl_utils.structs import BlockType
from tqdm import tqdm
from mineru.backend.hybrid.hybrid_model_output_to_middle_json import (
apply_server_side_postprocess,
append_page_model_list_to_middle_json,
finalize_middle_json,
init_middle_json,
@@ -722,6 +723,9 @@ def doc_analyze(
image_analysis: bool = True,
**kwargs,
):
client_side_output_generation = bool(
kwargs.pop("client_side_output_generation", False)
)
if predictor is None:
predictor = ModelSingleton().get_model(backend, model_path, server_url, **kwargs)
predictor = _maybe_enable_serial_execution(predictor, backend)
@@ -833,12 +837,20 @@ def doc_analyze(
f"speed: {round(len(model_list) / infer_time, 3)} page/s"
)
finalize_middle_json(
middle_json["pdf_info"],
hybrid_pipeline_model,
_ocr_enable,
_vlm_ocr_enable,
)
if client_side_output_generation:
apply_server_side_postprocess(
middle_json["pdf_info"],
hybrid_pipeline_model,
_ocr_enable,
_vlm_ocr_enable,
)
else:
finalize_middle_json(
middle_json["pdf_info"],
hybrid_pipeline_model,
_ocr_enable,
_vlm_ocr_enable,
)
close_pdfium_document(pdf_doc)
doc_closed = True
clean_memory(device)
@@ -861,6 +873,9 @@ async def aio_doc_analyze(
image_analysis: bool = True,
**kwargs,
):
client_side_output_generation = bool(
kwargs.pop("client_side_output_generation", False)
)
if predictor is None:
predictor = await _get_model_async(backend, model_path, server_url, **kwargs)
predictor = _maybe_enable_serial_execution(predictor, backend)
@@ -973,13 +988,21 @@ async def aio_doc_analyze(
f"speed: {round(len(model_list) / infer_time, 3)} page/s"
)
await asyncio.to_thread(
finalize_middle_json,
middle_json["pdf_info"],
hybrid_pipeline_model,
_ocr_enable,
_vlm_ocr_enable,
)
if client_side_output_generation:
apply_server_side_postprocess(
middle_json["pdf_info"],
hybrid_pipeline_model,
_ocr_enable,
_vlm_ocr_enable,
)
else:
await asyncio.to_thread(
finalize_middle_json,
middle_json["pdf_info"],
hybrid_pipeline_model,
_ocr_enable,
_vlm_ocr_enable,
)
close_pdfium_document(pdf_doc)
doc_closed = True
clean_memory(device)