From 61747bafddc6074d41a07e7ddc45f2e44b584ab7 Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Fri, 31 Oct 2025 14:57:17 +0800
Subject: [PATCH 01/28] fix: center-align column header for vlm accuracy in
 index.md table

---
 docs/en/quick_start/index.md | 2 +-
 docs/zh/quick_start/index.md | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/docs/en/quick_start/index.md b/docs/en/quick_start/index.md
index 8b77e986..cc349ba7 100644
--- a/docs/en/quick_start/index.md
+++ b/docs/en/quick_start/index.md
@@ -31,7 +31,7 @@ A WebUI developed based on Gradio, with a simple interface and only core parsing
         <tr>
             <th rowspan="2">Parsing Backend</th>
             <th rowspan="2">pipeline <br> (Accuracy<sup>1</sup> 82+)</th>
-            <th colspan="4">vlm (Accuracy<sup>1</sup> 90+)</th>
+            <th colspan="4" style="text-align:center;">vlm (Accuracy<sup>1</sup> 90+)</th>
         </tr>
         <tr>
             <th>transformers</th>
diff --git a/docs/zh/quick_start/index.md b/docs/zh/quick_start/index.md
index 82b432df..d00258b3 100644
--- a/docs/zh/quick_start/index.md
+++ b/docs/zh/quick_start/index.md
@@ -26,12 +26,12 @@
 >
 > 在非主线环境中，由于硬件、软件配置的多样性，以及第三方依赖项的兼容性问题，我们无法100%保证项目的完全可用性。因此，对于希望在非推荐环境中使用本项目的用户，我们建议先仔细阅读文档以及FAQ，大多数问题已经在FAQ中有对应的解决方案，除此之外我们鼓励社区反馈问题，以便我们能够逐步扩大支持范围。
 
-<table>
+<table border="1">
     <thead>
         <tr>
             <th rowspan="2">解析后端</th>
             <th rowspan="2">pipeline <br> (精度<sup>1</sup> 82+)</th>
-            <th colspan="4">vlm (精度<sup>1</sup> 90+)</th>
+            <th colspan="4" style="text-align:center;">vlm (精度<sup>1</sup> 90+)</th>
         </tr>
         <tr>
             <th>transformers</th>

From cce16daf1f5a63eb422b44ae4353b3d2919c6db4 Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Fri, 31 Oct 2025 15:37:08 +0800
Subject: [PATCH 02/28] fix: update JSON URL to point to the dev branch in
 configure_model function

---
 mineru/cli/models_download.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mineru/cli/models_download.py b/mineru/cli/models_download.py
index 2bb8922f..d01b335e 100644
--- a/mineru/cli/models_download.py
+++ b/mineru/cli/models_download.py
@@ -43,7 +43,7 @@ def download_and_modify_json(url, local_filename, modifications):
 
 def configure_model(model_dir, model_type):
     """配置模型"""
-    json_url = 'https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/mineru.template.json'
+    json_url = 'https://gcore.jsdelivr.net/gh/opendatalab/MinerU@dev/mineru.template.json'
     config_file_name = os.getenv('MINERU_TOOLS_CONFIG_JSON', 'mineru.json')
     home_dir = os.path.expanduser('~')
     config_file = os.path.join(home_dir, config_file_name)

From b614bef035e9acccbdced9f9c3e16da681cbb6ac Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Fri, 31 Oct 2025 17:50:59 +0800
Subject: [PATCH 03/28] feat: add multiprocessing support for PDF to image
 conversion with timeout handling

---
 mineru/utils/pdf_image_tools.py | 44 ++++++++++++++++++++++++++++++++-
 1 file changed, 43 insertions(+), 1 deletion(-)

diff --git a/mineru/utils/pdf_image_tools.py b/mineru/utils/pdf_image_tools.py
index adbb58dd..ca51c994 100644
--- a/mineru/utils/pdf_image_tools.py
+++ b/mineru/utils/pdf_image_tools.py
@@ -11,6 +11,8 @@ from mineru.utils.pdf_reader import image_to_b64str, image_to_bytes, page_to_ima
 from .enum_class import ImageType
 from .hash_utils import str_sha256
 
+from concurrent.futures import ProcessPoolExecutor, TimeoutError as FuturesTimeoutError
+
 
 def pdf_page_to_image(page: pdfium.PdfPage, dpi=200, image_type=ImageType.PIL) -> dict:
     """Convert pdfium.PdfDocument to image, Then convert the image to base64.
@@ -34,8 +36,48 @@ def pdf_page_to_image(page: pdfium.PdfPage, dpi=200, image_type=ImageType.PIL) -
 
     return image_dict
 
+#@todo 通过多进程渲染pdf提速
+def _load_images_from_pdf_worker(pdf_bytes, dpi, start_page_id, end_page_id, image_type):
+    """用于进程池的包装函数"""
+    return load_images_from_pdf_core(pdf_bytes, dpi, start_page_id, end_page_id, image_type)
+
 
 def load_images_from_pdf(
+        pdf_bytes: bytes,
+        dpi=200,
+        start_page_id=0,
+        end_page_id=None,
+        image_type=ImageType.PIL,
+        timeout=300,
+):
+    """带超时控制的 PDF 转图片函数
+
+    Args:
+        timeout (int): 超时时间(秒),默认 300 秒
+
+    Raises:
+        TimeoutError: 当转换超时时抛出
+    """
+    with ProcessPoolExecutor(max_workers=1) as executor:
+        future = executor.submit(
+            _load_images_from_pdf_worker,
+            pdf_bytes,
+            dpi,
+            start_page_id,
+            end_page_id,
+            image_type
+        )
+        try:
+            images_list = future.result(timeout=timeout)
+            pdf_doc = pdfium.PdfDocument(pdf_bytes)
+            return images_list, pdf_doc
+        except FuturesTimeoutError:
+            logger.error(f"PDF conversion timeout after {timeout}s")
+            executor.shutdown(wait=False, cancel_futures=True)
+            raise TimeoutError(f"PDF to images conversion timeout after {timeout}s")
+
+
+def load_images_from_pdf_core(
     pdf_bytes: bytes,
     dpi=200,
     start_page_id=0,
@@ -56,7 +98,7 @@ def load_images_from_pdf(
             image_dict = pdf_page_to_image(page, dpi=dpi, image_type=image_type)
             images_list.append(image_dict)
 
-    return images_list, pdf_doc
+    return images_list
 
 
 def cut_image(bbox: tuple, page_num: int, page_pil_img, return_path, image_writer: FileBasedDataWriter, scale=2):

From 305e3a61e8a4b824e2f56813cb62a8bc33ed7771 Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Sat, 1 Nov 2025 02:00:04 +0800
Subject: [PATCH 04/28] fix: disable tokenizers parallelism to prevent
 potential issues

---
 mineru/cli/common.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/mineru/cli/common.py b/mineru/cli/common.py
index 86c25fcb..ed63c2ac 100644
--- a/mineru/cli/common.py
+++ b/mineru/cli/common.py
@@ -20,6 +20,7 @@ from mineru.backend.vlm.vlm_analyze import aio_doc_analyze as aio_vlm_doc_analyz
 pdf_suffixes = ["pdf"]
 image_suffixes = ["png", "jpeg", "jp2", "webp", "gif", "bmp", "jpg", "tiff"]
 
+os.environ["TOKENIZERS_PARALLELISM"] = "false"
 
 def read_fn(path):
     if not isinstance(path, Path):

From 66d5f3dfd2fd75e482a0fe71aff590eb3d9b5f57 Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Mon, 3 Nov 2025 15:08:31 +0800
Subject: [PATCH 05/28] feat: refactor PDF image conversion to use
 get_end_page_id utility function and add multi-threading support

---
 mineru/cli/common.py            |  7 +--
 mineru/utils/pdf_image_tools.py | 82 +++++++++++++++++++++++++--------
 mineru/utils/pdf_page_id.py     | 10 ++++
 3 files changed, 75 insertions(+), 24 deletions(-)
 create mode 100644 mineru/utils/pdf_page_id.py

diff --git a/mineru/cli/common.py b/mineru/cli/common.py
index ed63c2ac..9dcedc64 100644
--- a/mineru/cli/common.py
+++ b/mineru/cli/common.py
@@ -16,6 +16,7 @@ from mineru.utils.pdf_image_tools import images_bytes_to_pdf_bytes
 from mineru.backend.vlm.vlm_middle_json_mkcontent import union_make as vlm_union_make
 from mineru.backend.vlm.vlm_analyze import doc_analyze as vlm_doc_analyze
 from mineru.backend.vlm.vlm_analyze import aio_doc_analyze as aio_vlm_doc_analyze
+from mineru.utils.pdf_page_id import get_end_page_id
 
 pdf_suffixes = ["pdf"]
 image_suffixes = ["png", "jpeg", "jp2", "webp", "gif", "bmp", "jpg", "tiff"]
@@ -49,11 +50,7 @@ def convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id=0, end_page
         # 从字节数据加载PDF
         pdf = pdfium.PdfDocument(pdf_bytes)
 
-        # 确定结束页
-        end_page_id = end_page_id if end_page_id is not None and end_page_id >= 0 else len(pdf) - 1
-        if end_page_id > len(pdf) - 1:
-            logger.warning("end_page_id is out of range, use pdf_docs length")
-            end_page_id = len(pdf) - 1
+        end_page_id = get_end_page_id(end_page_id, len(pdf))
 
         # 创建一个新的PDF文档
         output_pdf = pdfium.PdfDocument.new()
diff --git a/mineru/utils/pdf_image_tools.py b/mineru/utils/pdf_image_tools.py
index ca51c994..417e0206 100644
--- a/mineru/utils/pdf_image_tools.py
+++ b/mineru/utils/pdf_image_tools.py
@@ -8,8 +8,9 @@ from PIL import Image
 
 from mineru.data.data_reader_writer import FileBasedDataWriter
 from mineru.utils.pdf_reader import image_to_b64str, image_to_bytes, page_to_image
-from .enum_class import ImageType
-from .hash_utils import str_sha256
+from mineru.utils.enum_class import ImageType
+from mineru.utils.hash_utils import str_sha256
+from mineru.utils.pdf_page_id import get_end_page_id
 
 from concurrent.futures import ProcessPoolExecutor, TimeoutError as FuturesTimeoutError
 
@@ -36,7 +37,7 @@ def pdf_page_to_image(page: pdfium.PdfPage, dpi=200, image_type=ImageType.PIL) -
 
     return image_dict
 
-#@todo 通过多进程渲染pdf提速
+
 def _load_images_from_pdf_worker(pdf_bytes, dpi, start_page_id, end_page_id, image_type):
     """用于进程池的包装函数"""
     return load_images_from_pdf_core(pdf_bytes, dpi, start_page_id, end_page_id, image_type)
@@ -49,30 +50,74 @@ def load_images_from_pdf(
         end_page_id=None,
         image_type=ImageType.PIL,
         timeout=300,
+        threads=4,
 ):
-    """带超时控制的 PDF 转图片函数
+    """带超时控制的 PDF 转图片函数,支持多进程加速
 
     Args:
         timeout (int): 超时时间(秒),默认 300 秒
+        threads (int): 进程数,默认 4
 
     Raises:
         TimeoutError: 当转换超时时抛出
     """
-    with ProcessPoolExecutor(max_workers=1) as executor:
-        future = executor.submit(
-            _load_images_from_pdf_worker,
-            pdf_bytes,
-            dpi,
-            start_page_id,
-            end_page_id,
-            image_type
-        )
+    # 根据进程数调整超时时间
+    timeout = timeout // threads
+
+    pdf_doc = pdfium.PdfDocument(pdf_bytes)
+    end_page_id = get_end_page_id(end_page_id, len(pdf_doc))
+
+    # 计算总页数
+    total_pages = end_page_id - start_page_id + 1
+
+    # 实际使用的进程数不超过总页数
+    actual_threads = min(threads, total_pages)
+
+    # 根据实际进程数分组页面范围
+    pages_per_thread = max(1, total_pages // actual_threads)
+    page_ranges = []
+
+    for i in range(actual_threads):
+        range_start = start_page_id + i * pages_per_thread
+        if i == actual_threads - 1:
+            # 最后一个线程处理剩余所有页面
+            range_end = end_page_id
+        else:
+            range_end = min(start_page_id + (i + 1) * pages_per_thread - 1, end_page_id)
+
+        page_ranges.append((range_start, range_end))
+
+    with ProcessPoolExecutor(max_workers=actual_threads) as executor:
+        # 提交所有任务
+        futures = []
+        for range_start, range_end in page_ranges:
+            future = executor.submit(
+                _load_images_from_pdf_worker,
+                pdf_bytes,
+                dpi,
+                range_start,
+                range_end,
+                image_type
+            )
+            futures.append((range_start, future))
+
         try:
-            images_list = future.result(timeout=timeout)
-            pdf_doc = pdfium.PdfDocument(pdf_bytes)
+            # 收集结果并按页码排序
+            all_results = []
+            for range_start, future in futures:
+                images_list = future.result(timeout=timeout)
+                all_results.append((range_start, images_list))
+
+            # 按起始页码排序并合并结果
+            all_results.sort(key=lambda x: x[0])
+            images_list = []
+            for _, imgs in all_results:
+                images_list.extend(imgs)
+
             return images_list, pdf_doc
         except FuturesTimeoutError:
             logger.error(f"PDF conversion timeout after {timeout}s")
+            pdf_doc.close()
             executor.shutdown(wait=False, cancel_futures=True)
             raise TimeoutError(f"PDF to images conversion timeout after {timeout}s")
 
@@ -87,10 +132,7 @@ def load_images_from_pdf_core(
     images_list = []
     pdf_doc = pdfium.PdfDocument(pdf_bytes)
     pdf_page_num = len(pdf_doc)
-    end_page_id = end_page_id if end_page_id is not None and end_page_id >= 0 else pdf_page_num - 1
-    if end_page_id > pdf_page_num - 1:
-        logger.warning("end_page_id is out of range, use images length")
-        end_page_id = pdf_page_num - 1
+    end_page_id = get_end_page_id(end_page_id, pdf_page_num)
 
     for index in range(0, pdf_page_num):
         if start_page_id <= index <= end_page_id:
@@ -98,6 +140,8 @@ def load_images_from_pdf_core(
             image_dict = pdf_page_to_image(page, dpi=dpi, image_type=image_type)
             images_list.append(image_dict)
 
+    pdf_doc.close()
+
     return images_list
 
 
diff --git a/mineru/utils/pdf_page_id.py b/mineru/utils/pdf_page_id.py
new file mode 100644
index 00000000..3ad88993
--- /dev/null
+++ b/mineru/utils/pdf_page_id.py
@@ -0,0 +1,10 @@
+# Copyright (c) Opendatalab. All rights reserved.
+from loguru import logger
+
+
+def get_end_page_id(end_page_id, pdf_page_num):
+    end_page_id = end_page_id if end_page_id is not None and end_page_id >= 0 else pdf_page_num - 1
+    if end_page_id > pdf_page_num - 1:
+        logger.warning("end_page_id is out of range, use images length")
+        end_page_id = pdf_page_num - 1
+    return end_page_id
\ No newline at end of file

From 05e114f8b9312559e5afcc5e5784991bf13139eb Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Mon, 3 Nov 2025 15:32:40 +0800
Subject: [PATCH 06/28] feat: implement multiprocessing for PDF conversion to
 enhance performance

---
 mineru/cli/common.py | 18 ++++++++++++++----
 1 file changed, 14 insertions(+), 4 deletions(-)

diff --git a/mineru/cli/common.py b/mineru/cli/common.py
index 9dcedc64..84d71e4d 100644
--- a/mineru/cli/common.py
+++ b/mineru/cli/common.py
@@ -3,6 +3,7 @@ import io
 import json
 import os
 import copy
+from multiprocessing import Pool
 from pathlib import Path
 
 import pypdfium2 as pdfium
@@ -77,12 +78,21 @@ def convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id=0, end_page
     return output_bytes
 
 
+def _convert_pdf_in_process(args):
+    """在独立进程中执行PDF转换"""
+    pdf_bytes, start_page_id, end_page_id = args
+    return convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id, end_page_id)
+
+
 def _prepare_pdf_bytes(pdf_bytes_list, start_page_id, end_page_id):
     """准备处理PDF字节数据"""
-    result = []
-    for pdf_bytes in pdf_bytes_list:
-        new_pdf_bytes = convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id, end_page_id)
-        result.append(new_pdf_bytes)
+    # 准备参数列表
+    args_list = [(pdf_bytes, start_page_id, end_page_id) for pdf_bytes in pdf_bytes_list]
+
+    # 使用进程池执行转换
+    with Pool(processes=min(len(pdf_bytes_list), min(os.cpu_count() or 1, 4))) as pool:
+        result = pool.map(_convert_pdf_in_process, args_list)
+
     return result
 
 

From bffc6aff5387bdb9796de7ec2e94efc06a375397 Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Mon, 3 Nov 2025 15:44:33 +0800
Subject: [PATCH 07/28] fix: streamline PDF conversion process by restructuring
 try-except block and ensuring proper resource management

---
 mineru/cli/common.py | 13 ++++---------
 1 file changed, 4 insertions(+), 9 deletions(-)

diff --git a/mineru/cli/common.py b/mineru/cli/common.py
index 84d71e4d..ec84a4a2 100644
--- a/mineru/cli/common.py
+++ b/mineru/cli/common.py
@@ -47,15 +47,11 @@ def prepare_env(output_dir, pdf_file_name, parse_method):
 
 
 def convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id=0, end_page_id=None):
+    pdf = pdfium.PdfDocument(pdf_bytes)
+    output_pdf = pdfium.PdfDocument.new()
     try:
-        # 从字节数据加载PDF
-        pdf = pdfium.PdfDocument(pdf_bytes)
-
         end_page_id = get_end_page_id(end_page_id, len(pdf))
 
-        # 创建一个新的PDF文档
-        output_pdf = pdfium.PdfDocument.new()
-
         # 选择要导入的页面索引
         page_indices = list(range(start_page_id, end_page_id + 1))
 
@@ -68,13 +64,12 @@ def convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id=0, end_page
 
         # 获取字节数据
         output_bytes = output_buffer.getvalue()
-
-        pdf.close()  # 关闭原PDF文档以释放资源
-        output_pdf.close()  # 关闭新PDF文档以释放资源
     except Exception as e:
         logger.warning(f"Error in converting PDF bytes: {e}, Using original PDF bytes.")
         output_bytes = pdf_bytes
 
+    pdf.close()
+    output_pdf.close()
     return output_bytes
 
 

From 4214634de8d258a82d313da8d097d0da6e2e6f25 Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Mon, 3 Nov 2025 18:48:31 +0800
Subject: [PATCH 08/28] feat: add timing logs for PDF byte preparation to
 improve performance monitoring

---
 mineru/cli/common.py | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/mineru/cli/common.py b/mineru/cli/common.py
index ec84a4a2..eef7622d 100644
--- a/mineru/cli/common.py
+++ b/mineru/cli/common.py
@@ -3,6 +3,7 @@ import io
 import json
 import os
 import copy
+import time
 from multiprocessing import Pool
 from pathlib import Path
 
@@ -81,6 +82,7 @@ def _convert_pdf_in_process(args):
 
 def _prepare_pdf_bytes(pdf_bytes_list, start_page_id, end_page_id):
     """准备处理PDF字节数据"""
+    start_time = time.time()
     # 准备参数列表
     args_list = [(pdf_bytes, start_page_id, end_page_id) for pdf_bytes in pdf_bytes_list]
 
@@ -88,6 +90,15 @@ def _prepare_pdf_bytes(pdf_bytes_list, start_page_id, end_page_id):
     with Pool(processes=min(len(pdf_bytes_list), min(os.cpu_count() or 1, 4))) as pool:
         result = pool.map(_convert_pdf_in_process, args_list)
 
+    logger.debug(f"Prepare PDF bytes cost: {round(time.time() - start_time, 2)}s")
+
+    start_time = time.time()
+    result = []
+    for pdf_bytes in pdf_bytes_list:
+        new_pdf_bytes = convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id, end_page_id)
+        result.append(new_pdf_bytes)
+    logger.debug(f"Prepare PDF bytes cost: {round(time.time() - start_time, 2)}s")
+
     return result
 
 

From c32ff88400fdbf259560e03f14b0e11cf070aada Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Mon, 3 Nov 2025 19:07:19 +0800
Subject: [PATCH 09/28] refactor: rename check_mac_env to check_sys_env and add
 Windows environment detection

---
 mineru/cli/client.py                          |  2 +-
 mineru/cli/common.py                          | 26 +++++++++----------
 mineru/cli/gradio_app.py                      |  2 +-
 .../{check_mac_env.py => check_sys_env.py}    |  4 +++
 4 files changed, 19 insertions(+), 15 deletions(-)
 rename mineru/utils/{check_mac_env.py => check_sys_env.py} (92%)

diff --git a/mineru/cli/client.py b/mineru/cli/client.py
index fae1e888..a6b127e9 100644
--- a/mineru/cli/client.py
+++ b/mineru/cli/client.py
@@ -4,7 +4,7 @@ import click
 from pathlib import Path
 from loguru import logger
 
-from mineru.utils.check_mac_env import is_mac_os_version_supported
+from mineru.utils.check_sys_env import is_mac_os_version_supported
 from mineru.utils.cli_parser import arg_parse
 from mineru.utils.config_reader import get_device
 from mineru.utils.guess_suffix_or_lang import guess_suffix_by_path
diff --git a/mineru/cli/common.py b/mineru/cli/common.py
index eef7622d..78297aa3 100644
--- a/mineru/cli/common.py
+++ b/mineru/cli/common.py
@@ -11,6 +11,7 @@ import pypdfium2 as pdfium
 from loguru import logger
 
 from mineru.data.data_reader_writer import FileBasedDataWriter
+from mineru.utils.check_sys_env import is_windows_environment
 from mineru.utils.draw_bbox import draw_layout_bbox, draw_span_bbox, draw_line_sort_bbox
 from mineru.utils.enum_class import MakeMode
 from mineru.utils.guess_suffix_or_lang import guess_suffix_by_bytes
@@ -82,21 +83,20 @@ def _convert_pdf_in_process(args):
 
 def _prepare_pdf_bytes(pdf_bytes_list, start_page_id, end_page_id):
     """准备处理PDF字节数据"""
-    start_time = time.time()
-    # 准备参数列表
-    args_list = [(pdf_bytes, start_page_id, end_page_id) for pdf_bytes in pdf_bytes_list]
-
-    # 使用进程池执行转换
-    with Pool(processes=min(len(pdf_bytes_list), min(os.cpu_count() or 1, 4))) as pool:
-        result = pool.map(_convert_pdf_in_process, args_list)
-
-    logger.debug(f"Prepare PDF bytes cost: {round(time.time() - start_time, 2)}s")
-
     start_time = time.time()
     result = []
-    for pdf_bytes in pdf_bytes_list:
-        new_pdf_bytes = convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id, end_page_id)
-        result.append(new_pdf_bytes)
+    if is_windows_environment():
+        for pdf_bytes in pdf_bytes_list:
+            new_pdf_bytes = convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id, end_page_id)
+            result.append(new_pdf_bytes)
+    else:
+        # 准备参数列表
+        args_list = [(pdf_bytes, start_page_id, end_page_id) for pdf_bytes in pdf_bytes_list]
+
+        # 使用进程池执行转换
+        with Pool(processes=min(len(pdf_bytes_list), min(os.cpu_count() or 1, 4))) as pool:
+            result = pool.map(_convert_pdf_in_process, args_list)
+
     logger.debug(f"Prepare PDF bytes cost: {round(time.time() - start_time, 2)}s")
 
     return result
diff --git a/mineru/cli/gradio_app.py b/mineru/cli/gradio_app.py
index 4157d9d7..b29bf987 100644
--- a/mineru/cli/gradio_app.py
+++ b/mineru/cli/gradio_app.py
@@ -13,7 +13,7 @@ from gradio_pdf import PDF
 from loguru import logger
 
 from mineru.cli.common import prepare_env, read_fn, aio_do_parse, pdf_suffixes, image_suffixes
-from mineru.utils.check_mac_env import is_mac_os_version_supported
+from mineru.utils.check_sys_env import is_mac_os_version_supported
 from mineru.utils.cli_parser import arg_parse
 from mineru.utils.hash_utils import str_sha256
 
diff --git a/mineru/utils/check_mac_env.py b/mineru/utils/check_sys_env.py
similarity index 92%
rename from mineru/utils/check_mac_env.py
rename to mineru/utils/check_sys_env.py
index 41129c47..31fbfc34 100644
--- a/mineru/utils/check_mac_env.py
+++ b/mineru/utils/check_sys_env.py
@@ -4,6 +4,10 @@ import platform
 from packaging import version
 
 
+def is_windows_environment() -> bool:
+    return platform.system() == "Windows"
+
+
 # Detect if the current environment is a Mac computer
 def is_mac_environment() -> bool:
     return platform.system() == "Darwin"

From 4afa045545d0f2c4cfc72ec89adf100200de40a4 Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Mon, 3 Nov 2025 19:10:25 +0800
Subject: [PATCH 10/28] refactor: update import statement to use check_sys_env
 and adjust logging level for image loading

---
 mineru/backend/vlm/vlm_analyze.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/mineru/backend/vlm/vlm_analyze.py b/mineru/backend/vlm/vlm_analyze.py
index ad481753..a317f69e 100644
--- a/mineru/backend/vlm/vlm_analyze.py
+++ b/mineru/backend/vlm/vlm_analyze.py
@@ -8,7 +8,7 @@ from .utils import enable_custom_logits_processors, set_default_gpu_memory_utili
 from .model_output_to_middle_json import result_to_middle_json
 from ...data.data_reader_writer import DataWriter
 from mineru.utils.pdf_image_tools import load_images_from_pdf
-from ...utils.check_mac_env import is_mac_os_version_supported
+from ...utils.check_sys_env import is_mac_os_version_supported
 from ...utils.config_reader import get_device
 
 from ...utils.enum_class import ImageType
@@ -177,7 +177,7 @@ async def aio_doc_analyze(
     images_list, pdf_doc = load_images_from_pdf(pdf_bytes, image_type=ImageType.PIL)
     images_pil_list = [image_dict["img_pil"] for image_dict in images_list]
     # load_images_time = round(time.time() - load_images_start, 2)
-    # logger.info(f"load images cost: {load_images_time}, speed: {round(len(images_base64_list)/load_images_time, 3)} images/s")
+    # logger.debug(f"load images cost: {load_images_time}, speed: {round(len(images_pil_list)/load_images_time, 3)} images/s")
 
     # infer_start = time.time()
     results = await predictor.aio_batch_two_step_extract(images=images_pil_list)

From 245ae28c2723f3a0e3e55378c83c76954f094015 Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Mon, 3 Nov 2025 19:11:05 +0800
Subject: [PATCH 11/28] refactor: optimize page range calculation and enhance
 logging for image conversion process

---
 mineru/backend/pipeline/pipeline_analyze.py |  3 +++
 mineru/utils/pdf_image_tools.py             | 14 ++++++++------
 2 files changed, 11 insertions(+), 6 deletions(-)

diff --git a/mineru/backend/pipeline/pipeline_analyze.py b/mineru/backend/pipeline/pipeline_analyze.py
index 246d0111..e57da57b 100644
--- a/mineru/backend/pipeline/pipeline_analyze.py
+++ b/mineru/backend/pipeline/pipeline_analyze.py
@@ -99,7 +99,10 @@ def doc_analyze(
         _lang = lang_list[pdf_idx]
 
         # 收集每个数据集中的页面
+        # load_images_start = time.time()
         images_list, pdf_doc = load_images_from_pdf(pdf_bytes, image_type=ImageType.PIL)
+        # load_images_time = round(time.time() - load_images_start, 2)
+        # logger.debug(f"load images cost: {load_images_time}, speed: {round(len(images_list) / load_images_time, 3)} images/s")
         all_image_lists.append(images_list)
         all_pdf_docs.append(pdf_doc)
         for page_idx in range(len(images_list)):
diff --git a/mineru/utils/pdf_image_tools.py b/mineru/utils/pdf_image_tools.py
index 417e0206..865fa400 100644
--- a/mineru/utils/pdf_image_tools.py
+++ b/mineru/utils/pdf_image_tools.py
@@ -83,10 +83,12 @@ def load_images_from_pdf(
             # 最后一个线程处理剩余所有页面
             range_end = end_page_id
         else:
-            range_end = min(start_page_id + (i + 1) * pages_per_thread - 1, end_page_id)
+            range_end = start_page_id + (i + 1) * pages_per_thread - 1
 
         page_ranges.append((range_start, range_end))
 
+    # logger.debug(f"PDF to images using {actual_threads} processes, page ranges: {page_ranges}")
+
     with ProcessPoolExecutor(max_workers=actual_threads) as executor:
         # 提交所有任务
         futures = []
@@ -134,11 +136,11 @@ def load_images_from_pdf_core(
     pdf_page_num = len(pdf_doc)
     end_page_id = get_end_page_id(end_page_id, pdf_page_num)
 
-    for index in range(0, pdf_page_num):
-        if start_page_id <= index <= end_page_id:
-            page = pdf_doc[index]
-            image_dict = pdf_page_to_image(page, dpi=dpi, image_type=image_type)
-            images_list.append(image_dict)
+    for index in range(start_page_id, end_page_id+1):
+        # logger.debug(f"Converting page {index}/{pdf_page_num} to image")
+        page = pdf_doc[index]
+        image_dict = pdf_page_to_image(page, dpi=dpi, image_type=image_type)
+        images_list.append(image_dict)
 
     pdf_doc.close()
 

From 5999f6664f09d8f9d17d10e8930d51769ec88190 Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Mon, 3 Nov 2025 19:31:39 +0800
Subject: [PATCH 12/28] refactor: simplify PDF byte preparation by removing
 multiprocessing and enhancing direct conversion

---
 mineru/cli/common.py | 24 +++---------------------
 1 file changed, 3 insertions(+), 21 deletions(-)

diff --git a/mineru/cli/common.py b/mineru/cli/common.py
index 78297aa3..219f2a78 100644
--- a/mineru/cli/common.py
+++ b/mineru/cli/common.py
@@ -75,30 +75,12 @@ def convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id=0, end_page
     return output_bytes
 
 
-def _convert_pdf_in_process(args):
-    """在独立进程中执行PDF转换"""
-    pdf_bytes, start_page_id, end_page_id = args
-    return convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id, end_page_id)
-
-
 def _prepare_pdf_bytes(pdf_bytes_list, start_page_id, end_page_id):
     """准备处理PDF字节数据"""
-    start_time = time.time()
     result = []
-    if is_windows_environment():
-        for pdf_bytes in pdf_bytes_list:
-            new_pdf_bytes = convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id, end_page_id)
-            result.append(new_pdf_bytes)
-    else:
-        # 准备参数列表
-        args_list = [(pdf_bytes, start_page_id, end_page_id) for pdf_bytes in pdf_bytes_list]
-
-        # 使用进程池执行转换
-        with Pool(processes=min(len(pdf_bytes_list), min(os.cpu_count() or 1, 4))) as pool:
-            result = pool.map(_convert_pdf_in_process, args_list)
-
-    logger.debug(f"Prepare PDF bytes cost: {round(time.time() - start_time, 2)}s")
-
+    for pdf_bytes in pdf_bytes_list:
+        new_pdf_bytes = convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id, end_page_id)
+        result.append(new_pdf_bytes)
     return result
 
 

From 5349fd7ccdcf1f3ba4d3aaf62d7864c5ac2695b8 Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Mon, 3 Nov 2025 19:41:22 +0800
Subject: [PATCH 13/28] refactor: enhance PDF image loading by removing
 multiprocessing for Windows environment and improving logging

---
 mineru/backend/pipeline/pipeline_analyze.py |   6 +-
 mineru/utils/pdf_image_tools.py             | 111 +++++++++++---------
 2 files changed, 64 insertions(+), 53 deletions(-)

diff --git a/mineru/backend/pipeline/pipeline_analyze.py b/mineru/backend/pipeline/pipeline_analyze.py
index e57da57b..aa78007f 100644
--- a/mineru/backend/pipeline/pipeline_analyze.py
+++ b/mineru/backend/pipeline/pipeline_analyze.py
@@ -99,10 +99,10 @@ def doc_analyze(
         _lang = lang_list[pdf_idx]
 
         # 收集每个数据集中的页面
-        # load_images_start = time.time()
+        load_images_start = time.time()
         images_list, pdf_doc = load_images_from_pdf(pdf_bytes, image_type=ImageType.PIL)
-        # load_images_time = round(time.time() - load_images_start, 2)
-        # logger.debug(f"load images cost: {load_images_time}, speed: {round(len(images_list) / load_images_time, 3)} images/s")
+        load_images_time = round(time.time() - load_images_start, 2)
+        logger.debug(f"load images cost: {load_images_time}, speed: {round(len(images_list) / load_images_time, 3)} images/s")
         all_image_lists.append(images_list)
         all_pdf_docs.append(pdf_doc)
         for page_idx in range(len(images_list)):
diff --git a/mineru/utils/pdf_image_tools.py b/mineru/utils/pdf_image_tools.py
index 865fa400..32e0f831 100644
--- a/mineru/utils/pdf_image_tools.py
+++ b/mineru/utils/pdf_image_tools.py
@@ -7,6 +7,7 @@ from loguru import logger
 from PIL import Image
 
 from mineru.data.data_reader_writer import FileBasedDataWriter
+from mineru.utils.check_sys_env import is_windows_environment
 from mineru.utils.pdf_reader import image_to_b64str, image_to_bytes, page_to_image
 from mineru.utils.enum_class import ImageType
 from mineru.utils.hash_utils import str_sha256
@@ -61,67 +62,77 @@ def load_images_from_pdf(
     Raises:
         TimeoutError: 当转换超时时抛出
     """
-    # 根据进程数调整超时时间
-    timeout = timeout // threads
-
     pdf_doc = pdfium.PdfDocument(pdf_bytes)
-    end_page_id = get_end_page_id(end_page_id, len(pdf_doc))
+    if is_windows_environment():
+        # Windows 环境下不使用多进程
+        return load_images_from_pdf_core(
+            pdf_bytes,
+            dpi,
+            start_page_id,
+            end_page_id,
+            image_type
+        ), pdf_doc
+    else:
+        # 根据进程数调整超时时间
+        timeout = timeout // threads
 
-    # 计算总页数
-    total_pages = end_page_id - start_page_id + 1
+        end_page_id = get_end_page_id(end_page_id, len(pdf_doc))
 
-    # 实际使用的进程数不超过总页数
-    actual_threads = min(threads, total_pages)
+        # 计算总页数
+        total_pages = end_page_id - start_page_id + 1
 
-    # 根据实际进程数分组页面范围
-    pages_per_thread = max(1, total_pages // actual_threads)
-    page_ranges = []
+        # 实际使用的进程数不超过总页数
+        actual_threads = min(threads, total_pages)
 
-    for i in range(actual_threads):
-        range_start = start_page_id + i * pages_per_thread
-        if i == actual_threads - 1:
-            # 最后一个线程处理剩余所有页面
-            range_end = end_page_id
-        else:
-            range_end = start_page_id + (i + 1) * pages_per_thread - 1
+        # 根据实际进程数分组页面范围
+        pages_per_thread = max(1, total_pages // actual_threads)
+        page_ranges = []
 
-        page_ranges.append((range_start, range_end))
+        for i in range(actual_threads):
+            range_start = start_page_id + i * pages_per_thread
+            if i == actual_threads - 1:
+                # 最后一个线程处理剩余所有页面
+                range_end = end_page_id
+            else:
+                range_end = start_page_id + (i + 1) * pages_per_thread - 1
 
-    # logger.debug(f"PDF to images using {actual_threads} processes, page ranges: {page_ranges}")
+            page_ranges.append((range_start, range_end))
 
-    with ProcessPoolExecutor(max_workers=actual_threads) as executor:
-        # 提交所有任务
-        futures = []
-        for range_start, range_end in page_ranges:
-            future = executor.submit(
-                _load_images_from_pdf_worker,
-                pdf_bytes,
-                dpi,
-                range_start,
-                range_end,
-                image_type
-            )
-            futures.append((range_start, future))
+        # logger.debug(f"PDF to images using {actual_threads} processes, page ranges: {page_ranges}")
 
-        try:
-            # 收集结果并按页码排序
-            all_results = []
-            for range_start, future in futures:
-                images_list = future.result(timeout=timeout)
-                all_results.append((range_start, images_list))
+        with ProcessPoolExecutor(max_workers=actual_threads) as executor:
+            # 提交所有任务
+            futures = []
+            for range_start, range_end in page_ranges:
+                future = executor.submit(
+                    _load_images_from_pdf_worker,
+                    pdf_bytes,
+                    dpi,
+                    range_start,
+                    range_end,
+                    image_type
+                )
+                futures.append((range_start, future))
 
-            # 按起始页码排序并合并结果
-            all_results.sort(key=lambda x: x[0])
-            images_list = []
-            for _, imgs in all_results:
-                images_list.extend(imgs)
+            try:
+                # 收集结果并按页码排序
+                all_results = []
+                for range_start, future in futures:
+                    images_list = future.result(timeout=timeout)
+                    all_results.append((range_start, images_list))
 
-            return images_list, pdf_doc
-        except FuturesTimeoutError:
-            logger.error(f"PDF conversion timeout after {timeout}s")
-            pdf_doc.close()
-            executor.shutdown(wait=False, cancel_futures=True)
-            raise TimeoutError(f"PDF to images conversion timeout after {timeout}s")
+                # 按起始页码排序并合并结果
+                all_results.sort(key=lambda x: x[0])
+                images_list = []
+                for _, imgs in all_results:
+                    images_list.extend(imgs)
+
+                return images_list, pdf_doc
+            except FuturesTimeoutError:
+                logger.error(f"PDF conversion timeout after {timeout}s")
+                pdf_doc.close()
+                executor.shutdown(wait=False, cancel_futures=True)
+                raise TimeoutError(f"PDF to images conversion timeout after {timeout}s")
 
 
 def load_images_from_pdf_core(

From ace7f768696c1de7dcc61399f579357c49fef935 Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Mon, 3 Nov 2025 20:26:34 +0800
Subject: [PATCH 14/28] refactor: move PDF byte conversion functions to
 pdf_page_tools and simplify logic

---
 demo/demo.py                    |  3 ++-
 mineru/cli/common.py            | 34 +---------------------------
 mineru/utils/pdf_image_tools.py | 16 +++++++------
 mineru/utils/pdf_page_id.py     | 10 ---------
 mineru/utils/pdf_page_tools.py  | 40 +++++++++++++++++++++++++++++++++
 tests/unittest/test_e2e.py      |  2 +-
 6 files changed, 53 insertions(+), 52 deletions(-)
 delete mode 100644 mineru/utils/pdf_page_id.py
 create mode 100644 mineru/utils/pdf_page_tools.py

diff --git a/demo/demo.py b/demo/demo.py
index 5da0dbf1..f2e258da 100644
--- a/demo/demo.py
+++ b/demo/demo.py
@@ -6,7 +6,8 @@ from pathlib import Path
 
 from loguru import logger
 
-from mineru.cli.common import convert_pdf_bytes_to_bytes_by_pypdfium2, prepare_env, read_fn
+from mineru.cli.common import prepare_env, read_fn
+from mineru.utils.pdf_page_tools import convert_pdf_bytes_to_bytes_by_pypdfium2
 from mineru.data.data_reader_writer import FileBasedDataWriter
 from mineru.utils.draw_bbox import draw_layout_bbox, draw_span_bbox
 from mineru.utils.enum_class import MakeMode
diff --git a/mineru/cli/common.py b/mineru/cli/common.py
index 219f2a78..399eac79 100644
--- a/mineru/cli/common.py
+++ b/mineru/cli/common.py
@@ -1,17 +1,12 @@
 # Copyright (c) Opendatalab. All rights reserved.
-import io
 import json
 import os
 import copy
-import time
-from multiprocessing import Pool
 from pathlib import Path
 
-import pypdfium2 as pdfium
 from loguru import logger
 
 from mineru.data.data_reader_writer import FileBasedDataWriter
-from mineru.utils.check_sys_env import is_windows_environment
 from mineru.utils.draw_bbox import draw_layout_bbox, draw_span_bbox, draw_line_sort_bbox
 from mineru.utils.enum_class import MakeMode
 from mineru.utils.guess_suffix_or_lang import guess_suffix_by_bytes
@@ -19,7 +14,7 @@ from mineru.utils.pdf_image_tools import images_bytes_to_pdf_bytes
 from mineru.backend.vlm.vlm_middle_json_mkcontent import union_make as vlm_union_make
 from mineru.backend.vlm.vlm_analyze import doc_analyze as vlm_doc_analyze
 from mineru.backend.vlm.vlm_analyze import aio_doc_analyze as aio_vlm_doc_analyze
-from mineru.utils.pdf_page_id import get_end_page_id
+from mineru.utils.pdf_page_tools import convert_pdf_bytes_to_bytes_by_pypdfium2
 
 pdf_suffixes = ["pdf"]
 image_suffixes = ["png", "jpeg", "jp2", "webp", "gif", "bmp", "jpg", "tiff"]
@@ -48,33 +43,6 @@ def prepare_env(output_dir, pdf_file_name, parse_method):
     return local_image_dir, local_md_dir
 
 
-def convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id=0, end_page_id=None):
-    pdf = pdfium.PdfDocument(pdf_bytes)
-    output_pdf = pdfium.PdfDocument.new()
-    try:
-        end_page_id = get_end_page_id(end_page_id, len(pdf))
-
-        # 选择要导入的页面索引
-        page_indices = list(range(start_page_id, end_page_id + 1))
-
-        # 从原PDF导入页面到新PDF
-        output_pdf.import_pages(pdf, page_indices)
-
-        # 将新PDF保存到内存缓冲区
-        output_buffer = io.BytesIO()
-        output_pdf.save(output_buffer)
-
-        # 获取字节数据
-        output_bytes = output_buffer.getvalue()
-    except Exception as e:
-        logger.warning(f"Error in converting PDF bytes: {e}, Using original PDF bytes.")
-        output_bytes = pdf_bytes
-
-    pdf.close()
-    output_pdf.close()
-    return output_bytes
-
-
 def _prepare_pdf_bytes(pdf_bytes_list, start_page_id, end_page_id):
     """准备处理PDF字节数据"""
     result = []
diff --git a/mineru/utils/pdf_image_tools.py b/mineru/utils/pdf_image_tools.py
index 32e0f831..65d6c3f9 100644
--- a/mineru/utils/pdf_image_tools.py
+++ b/mineru/utils/pdf_image_tools.py
@@ -11,7 +11,7 @@ from mineru.utils.check_sys_env import is_windows_environment
 from mineru.utils.pdf_reader import image_to_b64str, image_to_bytes, page_to_image
 from mineru.utils.enum_class import ImageType
 from mineru.utils.hash_utils import str_sha256
-from mineru.utils.pdf_page_id import get_end_page_id
+from mineru.utils.pdf_page_tools import get_end_page_id, convert_pdf_bytes_to_bytes_by_pypdfium2
 
 from concurrent.futures import ProcessPoolExecutor, TimeoutError as FuturesTimeoutError
 
@@ -69,7 +69,7 @@ def load_images_from_pdf(
             pdf_bytes,
             dpi,
             start_page_id,
-            end_page_id,
+            get_end_page_id(end_page_id, len(pdf_doc)),
             image_type
         ), pdf_doc
     else:
@@ -143,13 +143,15 @@ def load_images_from_pdf_core(
     image_type=ImageType.PIL,  # PIL or BASE64
 ):
     images_list = []
+    pdf_bytes = convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id, end_page_id)
     pdf_doc = pdfium.PdfDocument(pdf_bytes)
-    pdf_page_num = len(pdf_doc)
-    end_page_id = get_end_page_id(end_page_id, pdf_page_num)
+    # pdf_page_num = len(pdf_doc)
+    # end_page_id = get_end_page_id(end_page_id, pdf_page_num)
 
-    for index in range(start_page_id, end_page_id+1):
-        # logger.debug(f"Converting page {index}/{pdf_page_num} to image")
-        page = pdf_doc[index]
+    # for index in range(start_page_id, end_page_id+1):
+    #     # logger.debug(f"Converting page {index}/{pdf_page_num} to image")
+    #     page = pdf_doc[index]
+    for page in pdf_doc:
         image_dict = pdf_page_to_image(page, dpi=dpi, image_type=image_type)
         images_list.append(image_dict)
 
diff --git a/mineru/utils/pdf_page_id.py b/mineru/utils/pdf_page_id.py
deleted file mode 100644
index 3ad88993..00000000
--- a/mineru/utils/pdf_page_id.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# Copyright (c) Opendatalab. All rights reserved.
-from loguru import logger
-
-
-def get_end_page_id(end_page_id, pdf_page_num):
-    end_page_id = end_page_id if end_page_id is not None and end_page_id >= 0 else pdf_page_num - 1
-    if end_page_id > pdf_page_num - 1:
-        logger.warning("end_page_id is out of range, use images length")
-        end_page_id = pdf_page_num - 1
-    return end_page_id
\ No newline at end of file
diff --git a/mineru/utils/pdf_page_tools.py b/mineru/utils/pdf_page_tools.py
new file mode 100644
index 00000000..9ea957a5
--- /dev/null
+++ b/mineru/utils/pdf_page_tools.py
@@ -0,0 +1,40 @@
+# Copyright (c) Opendatalab. All rights reserved.
+import io
+
+import pypdfium2 as pdfium
+from loguru import logger
+
+
+def get_end_page_id(end_page_id, pdf_page_num):
+    end_page_id = end_page_id if end_page_id is not None and end_page_id >= 0 else pdf_page_num - 1
+    if end_page_id > pdf_page_num - 1:
+        logger.warning("end_page_id is out of range, use images length")
+        end_page_id = pdf_page_num - 1
+    return end_page_id
+
+
+def convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id=0, end_page_id=None):
+    pdf = pdfium.PdfDocument(pdf_bytes)
+    output_pdf = pdfium.PdfDocument.new()
+    try:
+        end_page_id = get_end_page_id(end_page_id, len(pdf))
+
+        # 选择要导入的页面索引
+        page_indices = list(range(start_page_id, end_page_id + 1))
+
+        # 从原PDF导入页面到新PDF
+        output_pdf.import_pages(pdf, page_indices)
+
+        # 将新PDF保存到内存缓冲区
+        output_buffer = io.BytesIO()
+        output_pdf.save(output_buffer)
+
+        # 获取字节数据
+        output_bytes = output_buffer.getvalue()
+    except Exception as e:
+        logger.warning(f"Error in converting PDF bytes: {e}, Using original PDF bytes.")
+        output_bytes = pdf_bytes
+
+    pdf.close()
+    output_pdf.close()
+    return output_bytes
diff --git a/tests/unittest/test_e2e.py b/tests/unittest/test_e2e.py
index d50e69a2..d018f9dc 100644
--- a/tests/unittest/test_e2e.py
+++ b/tests/unittest/test_e2e.py
@@ -7,10 +7,10 @@ from loguru import logger
 from bs4 import BeautifulSoup
 from fuzzywuzzy import fuzz
 from mineru.cli.common import (
-    convert_pdf_bytes_to_bytes_by_pypdfium2,
     prepare_env,
     read_fn,
 )
+from mineru.utils.pdf_page_tools import convert_pdf_bytes_to_bytes_by_pypdfium2
 from mineru.data.data_reader_writer import FileBasedDataWriter
 from mineru.utils.enum_class import MakeMode
 from mineru.backend.vlm.vlm_analyze import doc_analyze as vlm_doc_analyze

From b4c57116c1da22acb5e0c2c141e11253d4c7a645 Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Mon, 3 Nov 2025 20:57:18 +0800
Subject: [PATCH 15/28] refactor: move PDF byte conversion logic to pdf_page_id
 and simplify image conversion process

---
 demo/demo.py                    |  3 +--
 mineru/cli/common.py            | 31 ++++++++++++++++++++++++-
 mineru/utils/pdf_image_tools.py | 14 +++++-------
 mineru/utils/pdf_page_id.py     | 10 +++++++++
 mineru/utils/pdf_page_tools.py  | 40 ---------------------------------
 tests/unittest/test_e2e.py      |  3 +--
 6 files changed, 48 insertions(+), 53 deletions(-)
 create mode 100644 mineru/utils/pdf_page_id.py
 delete mode 100644 mineru/utils/pdf_page_tools.py

diff --git a/demo/demo.py b/demo/demo.py
index f2e258da..c52018a6 100644
--- a/demo/demo.py
+++ b/demo/demo.py
@@ -6,8 +6,7 @@ from pathlib import Path
 
 from loguru import logger
 
-from mineru.cli.common import prepare_env, read_fn
-from mineru.utils.pdf_page_tools import convert_pdf_bytes_to_bytes_by_pypdfium2
+from mineru.cli.common import prepare_env, read_fn, convert_pdf_bytes_to_bytes_by_pypdfium2
 from mineru.data.data_reader_writer import FileBasedDataWriter
 from mineru.utils.draw_bbox import draw_layout_bbox, draw_span_bbox
 from mineru.utils.enum_class import MakeMode
diff --git a/mineru/cli/common.py b/mineru/cli/common.py
index 399eac79..445003f2 100644
--- a/mineru/cli/common.py
+++ b/mineru/cli/common.py
@@ -1,10 +1,12 @@
 # Copyright (c) Opendatalab. All rights reserved.
+import io
 import json
 import os
 import copy
 from pathlib import Path
 
 from loguru import logger
+import pypdfium2 as pdfium
 
 from mineru.data.data_reader_writer import FileBasedDataWriter
 from mineru.utils.draw_bbox import draw_layout_bbox, draw_span_bbox, draw_line_sort_bbox
@@ -14,7 +16,7 @@ from mineru.utils.pdf_image_tools import images_bytes_to_pdf_bytes
 from mineru.backend.vlm.vlm_middle_json_mkcontent import union_make as vlm_union_make
 from mineru.backend.vlm.vlm_analyze import doc_analyze as vlm_doc_analyze
 from mineru.backend.vlm.vlm_analyze import aio_doc_analyze as aio_vlm_doc_analyze
-from mineru.utils.pdf_page_tools import convert_pdf_bytes_to_bytes_by_pypdfium2
+from mineru.utils.pdf_page_id import get_end_page_id
 
 pdf_suffixes = ["pdf"]
 image_suffixes = ["png", "jpeg", "jp2", "webp", "gif", "bmp", "jpg", "tiff"]
@@ -43,6 +45,33 @@ def prepare_env(output_dir, pdf_file_name, parse_method):
     return local_image_dir, local_md_dir
 
 
+def convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id=0, end_page_id=None):
+    pdf = pdfium.PdfDocument(pdf_bytes)
+    output_pdf = pdfium.PdfDocument.new()
+    try:
+        end_page_id = get_end_page_id(end_page_id, len(pdf))
+
+        # 选择要导入的页面索引
+        page_indices = list(range(start_page_id, end_page_id + 1))
+
+        # 从原PDF导入页面到新PDF
+        output_pdf.import_pages(pdf, page_indices)
+
+        # 将新PDF保存到内存缓冲区
+        output_buffer = io.BytesIO()
+        output_pdf.save(output_buffer)
+
+        # 获取字节数据
+        output_bytes = output_buffer.getvalue()
+    except Exception as e:
+        logger.warning(f"Error in converting PDF bytes: {e}, Using original PDF bytes.")
+        output_bytes = pdf_bytes
+
+    pdf.close()
+    output_pdf.close()
+    return output_bytes
+
+
 def _prepare_pdf_bytes(pdf_bytes_list, start_page_id, end_page_id):
     """准备处理PDF字节数据"""
     result = []
diff --git a/mineru/utils/pdf_image_tools.py b/mineru/utils/pdf_image_tools.py
index 65d6c3f9..f59877f9 100644
--- a/mineru/utils/pdf_image_tools.py
+++ b/mineru/utils/pdf_image_tools.py
@@ -11,7 +11,7 @@ from mineru.utils.check_sys_env import is_windows_environment
 from mineru.utils.pdf_reader import image_to_b64str, image_to_bytes, page_to_image
 from mineru.utils.enum_class import ImageType
 from mineru.utils.hash_utils import str_sha256
-from mineru.utils.pdf_page_tools import get_end_page_id, convert_pdf_bytes_to_bytes_by_pypdfium2
+from mineru.utils.pdf_page_id import get_end_page_id
 
 from concurrent.futures import ProcessPoolExecutor, TimeoutError as FuturesTimeoutError
 
@@ -143,15 +143,13 @@ def load_images_from_pdf_core(
     image_type=ImageType.PIL,  # PIL or BASE64
 ):
     images_list = []
-    pdf_bytes = convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id, end_page_id)
     pdf_doc = pdfium.PdfDocument(pdf_bytes)
-    # pdf_page_num = len(pdf_doc)
-    # end_page_id = get_end_page_id(end_page_id, pdf_page_num)
+    pdf_page_num = len(pdf_doc)
+    end_page_id = get_end_page_id(end_page_id, pdf_page_num)
 
-    # for index in range(start_page_id, end_page_id+1):
-    #     # logger.debug(f"Converting page {index}/{pdf_page_num} to image")
-    #     page = pdf_doc[index]
-    for page in pdf_doc:
+    for index in range(start_page_id, end_page_id+1):
+        # logger.debug(f"Converting page {index}/{pdf_page_num} to image")
+        page = pdf_doc[index]
         image_dict = pdf_page_to_image(page, dpi=dpi, image_type=image_type)
         images_list.append(image_dict)
 
diff --git a/mineru/utils/pdf_page_id.py b/mineru/utils/pdf_page_id.py
new file mode 100644
index 00000000..c1e72336
--- /dev/null
+++ b/mineru/utils/pdf_page_id.py
@@ -0,0 +1,10 @@
+# Copyright (c) Opendatalab. All rights reserved.
+from loguru import logger
+
+
+def get_end_page_id(end_page_id, pdf_page_num):
+    end_page_id = end_page_id if end_page_id is not None and end_page_id >= 0 else pdf_page_num - 1
+    if end_page_id > pdf_page_num - 1:
+        logger.warning("end_page_id is out of range, use images length")
+        end_page_id = pdf_page_num - 1
+    return end_page_id
diff --git a/mineru/utils/pdf_page_tools.py b/mineru/utils/pdf_page_tools.py
deleted file mode 100644
index 9ea957a5..00000000
--- a/mineru/utils/pdf_page_tools.py
+++ /dev/null
@@ -1,40 +0,0 @@
-# Copyright (c) Opendatalab. All rights reserved.
-import io
-
-import pypdfium2 as pdfium
-from loguru import logger
-
-
-def get_end_page_id(end_page_id, pdf_page_num):
-    end_page_id = end_page_id if end_page_id is not None and end_page_id >= 0 else pdf_page_num - 1
-    if end_page_id > pdf_page_num - 1:
-        logger.warning("end_page_id is out of range, use images length")
-        end_page_id = pdf_page_num - 1
-    return end_page_id
-
-
-def convert_pdf_bytes_to_bytes_by_pypdfium2(pdf_bytes, start_page_id=0, end_page_id=None):
-    pdf = pdfium.PdfDocument(pdf_bytes)
-    output_pdf = pdfium.PdfDocument.new()
-    try:
-        end_page_id = get_end_page_id(end_page_id, len(pdf))
-
-        # 选择要导入的页面索引
-        page_indices = list(range(start_page_id, end_page_id + 1))
-
-        # 从原PDF导入页面到新PDF
-        output_pdf.import_pages(pdf, page_indices)
-
-        # 将新PDF保存到内存缓冲区
-        output_buffer = io.BytesIO()
-        output_pdf.save(output_buffer)
-
-        # 获取字节数据
-        output_bytes = output_buffer.getvalue()
-    except Exception as e:
-        logger.warning(f"Error in converting PDF bytes: {e}, Using original PDF bytes.")
-        output_bytes = pdf_bytes
-
-    pdf.close()
-    output_pdf.close()
-    return output_bytes
diff --git a/tests/unittest/test_e2e.py b/tests/unittest/test_e2e.py
index d018f9dc..6457508f 100644
--- a/tests/unittest/test_e2e.py
+++ b/tests/unittest/test_e2e.py
@@ -8,9 +8,8 @@ from bs4 import BeautifulSoup
 from fuzzywuzzy import fuzz
 from mineru.cli.common import (
     prepare_env,
-    read_fn,
+    read_fn, convert_pdf_bytes_to_bytes_by_pypdfium2,
 )
-from mineru.utils.pdf_page_tools import convert_pdf_bytes_to_bytes_by_pypdfium2
 from mineru.data.data_reader_writer import FileBasedDataWriter
 from mineru.utils.enum_class import MakeMode
 from mineru.backend.vlm.vlm_analyze import doc_analyze as vlm_doc_analyze

From 20793957746a86338b3b8ccccac932726693ca6c Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Mon, 3 Nov 2025 21:02:58 +0800
Subject: [PATCH 16/28] refactor: adjust thread count based on CPU cores and
 comment out image loading time logging

---
 mineru/backend/pipeline/pipeline_analyze.py | 6 +++---
 mineru/utils/pdf_image_tools.py             | 2 ++
 2 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/mineru/backend/pipeline/pipeline_analyze.py b/mineru/backend/pipeline/pipeline_analyze.py
index aa78007f..e57da57b 100644
--- a/mineru/backend/pipeline/pipeline_analyze.py
+++ b/mineru/backend/pipeline/pipeline_analyze.py
@@ -99,10 +99,10 @@ def doc_analyze(
         _lang = lang_list[pdf_idx]
 
         # 收集每个数据集中的页面
-        load_images_start = time.time()
+        # load_images_start = time.time()
         images_list, pdf_doc = load_images_from_pdf(pdf_bytes, image_type=ImageType.PIL)
-        load_images_time = round(time.time() - load_images_start, 2)
-        logger.debug(f"load images cost: {load_images_time}, speed: {round(len(images_list) / load_images_time, 3)} images/s")
+        # load_images_time = round(time.time() - load_images_start, 2)
+        # logger.debug(f"load images cost: {load_images_time}, speed: {round(len(images_list) / load_images_time, 3)} images/s")
         all_image_lists.append(images_list)
         all_pdf_docs.append(pdf_doc)
         for page_idx in range(len(images_list)):
diff --git a/mineru/utils/pdf_image_tools.py b/mineru/utils/pdf_image_tools.py
index f59877f9..15e74128 100644
--- a/mineru/utils/pdf_image_tools.py
+++ b/mineru/utils/pdf_image_tools.py
@@ -1,4 +1,5 @@
 # Copyright (c) Opendatalab. All rights reserved.
+import os
 from io import BytesIO
 
 import numpy as np
@@ -74,6 +75,7 @@ def load_images_from_pdf(
         ), pdf_doc
     else:
         # 根据进程数调整超时时间
+        threads = min(os.cpu_count(), threads)
         timeout = timeout // threads
 
         end_page_id = get_end_page_id(end_page_id, len(pdf_doc))

From 2f120db20e8f7358dba62616181338c2d22bf95d Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Mon, 3 Nov 2025 21:24:13 +0800
Subject: [PATCH 17/28] fix: update JSON URL to point to the master branch for
 model configuration

---
 mineru/cli/models_download.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mineru/cli/models_download.py b/mineru/cli/models_download.py
index d01b335e..2bb8922f 100644
--- a/mineru/cli/models_download.py
+++ b/mineru/cli/models_download.py
@@ -43,7 +43,7 @@ def download_and_modify_json(url, local_filename, modifications):
 
 def configure_model(model_dir, model_type):
     """配置模型"""
-    json_url = 'https://gcore.jsdelivr.net/gh/opendatalab/MinerU@dev/mineru.template.json'
+    json_url = 'https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/mineru.template.json'
     config_file_name = os.getenv('MINERU_TOOLS_CONFIG_JSON', 'mineru.json')
     home_dir = os.path.expanduser('~')
     config_file = os.path.join(home_dir, config_file_name)

From 54417a51f86f6a4792a743247c1fe9a3e48f2c4f Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Mon, 3 Nov 2025 21:27:00 +0800
Subject: [PATCH 18/28] refactor: reorder import statements for clarity and
 consistency

---
 demo/demo.py               | 2 +-
 tests/unittest/test_e2e.py | 3 ++-
 2 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/demo/demo.py b/demo/demo.py
index c52018a6..5da0dbf1 100644
--- a/demo/demo.py
+++ b/demo/demo.py
@@ -6,7 +6,7 @@ from pathlib import Path
 
 from loguru import logger
 
-from mineru.cli.common import prepare_env, read_fn, convert_pdf_bytes_to_bytes_by_pypdfium2
+from mineru.cli.common import convert_pdf_bytes_to_bytes_by_pypdfium2, prepare_env, read_fn
 from mineru.data.data_reader_writer import FileBasedDataWriter
 from mineru.utils.draw_bbox import draw_layout_bbox, draw_span_bbox
 from mineru.utils.enum_class import MakeMode
diff --git a/tests/unittest/test_e2e.py b/tests/unittest/test_e2e.py
index 6457508f..d50e69a2 100644
--- a/tests/unittest/test_e2e.py
+++ b/tests/unittest/test_e2e.py
@@ -7,8 +7,9 @@ from loguru import logger
 from bs4 import BeautifulSoup
 from fuzzywuzzy import fuzz
 from mineru.cli.common import (
+    convert_pdf_bytes_to_bytes_by_pypdfium2,
     prepare_env,
-    read_fn, convert_pdf_bytes_to_bytes_by_pypdfium2,
+    read_fn,
 )
 from mineru.data.data_reader_writer import FileBasedDataWriter
 from mineru.utils.enum_class import MakeMode

From 6250c453d92f27434d302a0a1879f076c8142392 Mon Sep 17 00:00:00 2001
From: Xiaomeng Zhao <moe@myhloli.com>
Date: Mon, 3 Nov 2025 22:04:49 +0800
Subject: [PATCH 19/28] Update mineru/utils/pdf_image_tools.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 mineru/utils/pdf_image_tools.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mineru/utils/pdf_image_tools.py b/mineru/utils/pdf_image_tools.py
index 15e74128..a7203d2d 100644
--- a/mineru/utils/pdf_image_tools.py
+++ b/mineru/utils/pdf_image_tools.py
@@ -149,7 +149,7 @@ def load_images_from_pdf_core(
     pdf_page_num = len(pdf_doc)
     end_page_id = get_end_page_id(end_page_id, pdf_page_num)
 
-    for index in range(start_page_id, end_page_id+1):
+    for index in range(start_page_id, end_page_id + 1):
         # logger.debug(f"Converting page {index}/{pdf_page_num} to image")
         page = pdf_doc[index]
         image_dict = pdf_page_to_image(page, dpi=dpi, image_type=image_type)

From 74de2725cb6ea22e87fa5aca68e6907a64fcad13 Mon Sep 17 00:00:00 2001
From: Xiaomeng Zhao <moe@myhloli.com>
Date: Mon, 3 Nov 2025 22:08:20 +0800
Subject: [PATCH 20/28] Update mineru/utils/pdf_image_tools.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 mineru/utils/pdf_image_tools.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mineru/utils/pdf_image_tools.py b/mineru/utils/pdf_image_tools.py
index a7203d2d..fb06db80 100644
--- a/mineru/utils/pdf_image_tools.py
+++ b/mineru/utils/pdf_image_tools.py
@@ -75,7 +75,7 @@ def load_images_from_pdf(
         ), pdf_doc
     else:
         # 根据进程数调整超时时间
-        threads = min(os.cpu_count(), threads)
+        threads = min(os.cpu_count() or 1, threads)
         timeout = timeout // threads
 
         end_page_id = get_end_page_id(end_page_id, len(pdf_doc))

From a9c9501af6c1901ded3d6c12863fee98b02254c9 Mon Sep 17 00:00:00 2001
From: Xiaomeng Zhao <moe@myhloli.com>
Date: Mon, 3 Nov 2025 22:09:29 +0800
Subject: [PATCH 21/28] Update mineru/utils/pdf_image_tools.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 mineru/utils/pdf_image_tools.py | 2 --
 1 file changed, 2 deletions(-)

diff --git a/mineru/utils/pdf_image_tools.py b/mineru/utils/pdf_image_tools.py
index fb06db80..c98e3134 100644
--- a/mineru/utils/pdf_image_tools.py
+++ b/mineru/utils/pdf_image_tools.py
@@ -76,8 +76,6 @@ def load_images_from_pdf(
     else:
         # 根据进程数调整超时时间
         threads = min(os.cpu_count() or 1, threads)
-        timeout = timeout // threads
-
         end_page_id = get_end_page_id(end_page_id, len(pdf_doc))
 
         # 计算总页数

From 51df4d850891570737c805a945dd0e3e1dce4a93 Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Tue, 4 Nov 2025 09:54:45 +0800
Subject: [PATCH 22/28] refactor: enhance PDF conversion function parameters
 and improve thread handling logic

---
 mineru/utils/pdf_image_tools.py | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/mineru/utils/pdf_image_tools.py b/mineru/utils/pdf_image_tools.py
index c98e3134..5bbaeadb 100644
--- a/mineru/utils/pdf_image_tools.py
+++ b/mineru/utils/pdf_image_tools.py
@@ -57,6 +57,11 @@ def load_images_from_pdf(
     """带超时控制的 PDF 转图片函数,支持多进程加速
 
     Args:
+        pdf_bytes (bytes): PDF 文件的 bytes
+        dpi (int, optional): reset the dpi of dpi. Defaults to 200.
+        start_page_id (int, optional): 起始页码. Defaults to 0.
+        end_page_id (int | None, optional): 结束页码. Defaults to None.
+        image_type (ImageType, optional): 图片类型. Defaults to ImageType.PIL.
         timeout (int): 超时时间(秒),默认 300 秒
         threads (int): 进程数,默认 4
 
@@ -74,15 +79,13 @@ def load_images_from_pdf(
             image_type
         ), pdf_doc
     else:
-        # 根据进程数调整超时时间
-        threads = min(os.cpu_count() or 1, threads)
         end_page_id = get_end_page_id(end_page_id, len(pdf_doc))
 
         # 计算总页数
         total_pages = end_page_id - start_page_id + 1
 
         # 实际使用的进程数不超过总页数
-        actual_threads = min(threads, total_pages)
+        actual_threads = min(min(os.cpu_count() or 1, threads), total_pages)
 
         # 根据实际进程数分组页面范围
         pages_per_thread = max(1, total_pages // actual_threads)
@@ -129,7 +132,6 @@ def load_images_from_pdf(
 
                 return images_list, pdf_doc
             except FuturesTimeoutError:
-                logger.error(f"PDF conversion timeout after {timeout}s")
                 pdf_doc.close()
                 executor.shutdown(wait=False, cancel_futures=True)
                 raise TimeoutError(f"PDF to images conversion timeout after {timeout}s")

From be2369bdd417b2a799b89dc9175f2bc15dff287d Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Tue, 4 Nov 2025 19:09:33 +0800
Subject: [PATCH 23/28] feat: add ONNX configuration for thread management and
 integrate into table structure

---
 .../table/rec/slanet_plus/table_structure.py  |  4 ++++
 .../rec/unet_table/table_structure_unet.py    |  5 ++++
 mineru/model/utils/onnx_config.py             | 24 +++++++++++++++++++
 3 files changed, 33 insertions(+)
 create mode 100644 mineru/model/utils/onnx_config.py

diff --git a/mineru/model/table/rec/slanet_plus/table_structure.py b/mineru/model/table/rec/slanet_plus/table_structure.py
index c36c61db..99c704bc 100644
--- a/mineru/model/table/rec/slanet_plus/table_structure.py
+++ b/mineru/model/table/rec/slanet_plus/table_structure.py
@@ -16,6 +16,7 @@ from typing import Any, Dict, List, Tuple
 
 import numpy as np
 
+from mineru.model.utils.onnx_config import get_op_num_threads
 from .table_structure_utils import (
     OrtInferSession,
     TableLabelDecode,
@@ -29,6 +30,9 @@ class TableStructurer:
         self.preprocess_op = TablePreprocess()
         self.batch_preprocess_op = BatchTablePreprocess()
 
+        config["intra_op_num_threads"] = get_op_num_threads("MINERU_INTRA_OP_NUM_THREADS")
+        config["inter_op_num_threads"] = get_op_num_threads("MINERU_INTER_OP_NUM_THREADS")
+
         self.session = OrtInferSession(config)
 
         self.character = self.session.get_metadata()
diff --git a/mineru/model/table/rec/unet_table/table_structure_unet.py b/mineru/model/table/rec/unet_table/table_structure_unet.py
index 9c6b9826..642dee2d 100644
--- a/mineru/model/table/rec/unet_table/table_structure_unet.py
+++ b/mineru/model/table/rec/unet_table/table_structure_unet.py
@@ -5,6 +5,8 @@ from typing import Optional, Dict, Any, Tuple
 import cv2
 import numpy as np
 from skimage import measure
+
+from mineru.model.utils.onnx_config import get_op_num_threads
 from .utils import OrtInferSession, resize_img
 from .utils_table_line_rec import (
     get_table_line,
@@ -28,6 +30,9 @@ class TSRUnet:
         self.inp_height = 1024
         self.inp_width = 1024
 
+        config["intra_op_num_threads"] = get_op_num_threads("MINERU_INTRA_OP_NUM_THREADS")
+        config["inter_op_num_threads"] = get_op_num_threads("MINERU_INTER_OP_NUM_THREADS")
+
         self.session = OrtInferSession(config)
 
     def __call__(
diff --git a/mineru/model/utils/onnx_config.py b/mineru/model/utils/onnx_config.py
new file mode 100644
index 00000000..492ea743
--- /dev/null
+++ b/mineru/model/utils/onnx_config.py
@@ -0,0 +1,24 @@
+import os
+
+
+def get_op_num_threads(env_name: str) -> int:
+    env_value = os.getenv(env_name, None)
+    return get_op_num_threads_from_value(env_value)
+
+
+def get_op_num_threads_from_value(env_value: str) -> int:
+    if env_value is not None:
+        try:
+            num_threads = int(env_value)
+            if num_threads > 0:
+                return num_threads
+        except ValueError:
+            return -1
+    return -1
+
+
+if __name__ == '__main__':
+    print(get_op_num_threads_from_value('1'))
+    print(get_op_num_threads_from_value('0'))
+    print(get_op_num_threads_from_value('-1'))
+    print(get_op_num_threads_from_value('abc'))
\ No newline at end of file

From 5de8f1a19f06551a4549806a5621ebd5dddaff1f Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Tue, 4 Nov 2025 19:47:59 +0800
Subject: [PATCH 24/28] feat: add environment variables for PDF rendering
 timeout and ONNX thread management

---
 docs/en/usage/cli_tools.md                    | 11 +++++++
 docs/zh/usage/cli_tools.md                    | 12 ++++++++
 .../table/rec/slanet_plus/table_structure.py  |  2 +-
 .../rec/unet_table/table_structure_unet.py    |  2 +-
 mineru/model/utils/onnx_config.py             | 24 ---------------
 mineru/utils/os_env_config.py                 | 30 +++++++++++++++++++
 mineru/utils/pdf_image_tools.py               |  5 +++-
 7 files changed, 59 insertions(+), 27 deletions(-)
 delete mode 100644 mineru/model/utils/onnx_config.py
 create mode 100644 mineru/utils/os_env_config.py

diff --git a/docs/en/usage/cli_tools.md b/docs/en/usage/cli_tools.md
index 14d81408..5b81027e 100644
--- a/docs/en/usage/cli_tools.md
+++ b/docs/en/usage/cli_tools.md
@@ -100,3 +100,14 @@ Here are the environment variables and their descriptions:
     * Used to enable table merging functionality
     * Default is `true`, can be set to `false` via environment variable to disable table merging functionality.
 
+- `MINERU_PDF_LOAD_IMAGES_TIMEOUT`:
+    * Used to set the timeout period (in seconds) for rendering PDF to images
+    * Default is `300` seconds, can be set to other values via environment variable to adjust the image rendering timeout.
+
+- `MINERU_INTRA_OP_NUM_THREADS`:
+    * Used to set the intra_op thread count for ONNX models, affects the computation speed of individual operators
+    * Default is `-1` (auto-select), can be set to other values via environment variable to adjust the thread count.
+
+- `MINERU_INTER_OP_NUM_THREADS`:
+    * Used to set the inter_op thread count for ONNX models, affects the parallel execution of multiple operators
+    * Default is `-1` (auto-select), can be set to other values via environment variable to adjust the thread count.
diff --git a/docs/zh/usage/cli_tools.md b/docs/zh/usage/cli_tools.md
index 99f0c370..15d4d2eb 100644
--- a/docs/zh/usage/cli_tools.md
+++ b/docs/zh/usage/cli_tools.md
@@ -94,3 +94,15 @@ MinerU命令行工具的某些参数存在相同功能的环境变量配置，
 - `MINERU_TABLE_MERGE_ENABLE`：
     * 用于启用表格合并功能
     * 默认为`true`，可通过环境变量设置为`false`来禁用表格合并功能。
+
+- `MINERU_PDF_LOAD_IMAGES_TIMEOUT`：
+    * 用于设置将PDF渲染为图片的超时时间（秒）
+    * 默认为`300`秒，可通过环境变量设置为其他值以调整渲染图片的超时时间。
+
+- `MINERU_INTRA_OP_NUM_THREADS`：
+    * 用于设置onnx模型的intra_op线程数，影响单个算子的计算速度
+    * 默认为`-1`（自动选择），可通过环境变量设置为其他值以调整线程数。
+
+- `MINERU_INTER_OP_NUM_THREADS`：
+    * 用于设置onnx模型的inter_op线程数，影响多个算子的并行执行
+    * 默认为`-1`（自动选择），可通过环境变量设置为其他值以调整线程数。
diff --git a/mineru/model/table/rec/slanet_plus/table_structure.py b/mineru/model/table/rec/slanet_plus/table_structure.py
index 99c704bc..d9b2f0ce 100644
--- a/mineru/model/table/rec/slanet_plus/table_structure.py
+++ b/mineru/model/table/rec/slanet_plus/table_structure.py
@@ -16,7 +16,7 @@ from typing import Any, Dict, List, Tuple
 
 import numpy as np
 
-from mineru.model.utils.onnx_config import get_op_num_threads
+from mineru.utils.os_env_config import get_op_num_threads
 from .table_structure_utils import (
     OrtInferSession,
     TableLabelDecode,
diff --git a/mineru/model/table/rec/unet_table/table_structure_unet.py b/mineru/model/table/rec/unet_table/table_structure_unet.py
index 642dee2d..d20ced76 100644
--- a/mineru/model/table/rec/unet_table/table_structure_unet.py
+++ b/mineru/model/table/rec/unet_table/table_structure_unet.py
@@ -6,7 +6,7 @@ import cv2
 import numpy as np
 from skimage import measure
 
-from mineru.model.utils.onnx_config import get_op_num_threads
+from mineru.utils.os_env_config import get_op_num_threads
 from .utils import OrtInferSession, resize_img
 from .utils_table_line_rec import (
     get_table_line,
diff --git a/mineru/model/utils/onnx_config.py b/mineru/model/utils/onnx_config.py
deleted file mode 100644
index 492ea743..00000000
--- a/mineru/model/utils/onnx_config.py
+++ /dev/null
@@ -1,24 +0,0 @@
-import os
-
-
-def get_op_num_threads(env_name: str) -> int:
-    env_value = os.getenv(env_name, None)
-    return get_op_num_threads_from_value(env_value)
-
-
-def get_op_num_threads_from_value(env_value: str) -> int:
-    if env_value is not None:
-        try:
-            num_threads = int(env_value)
-            if num_threads > 0:
-                return num_threads
-        except ValueError:
-            return -1
-    return -1
-
-
-if __name__ == '__main__':
-    print(get_op_num_threads_from_value('1'))
-    print(get_op_num_threads_from_value('0'))
-    print(get_op_num_threads_from_value('-1'))
-    print(get_op_num_threads_from_value('abc'))
\ No newline at end of file
diff --git a/mineru/utils/os_env_config.py b/mineru/utils/os_env_config.py
new file mode 100644
index 00000000..684976ca
--- /dev/null
+++ b/mineru/utils/os_env_config.py
@@ -0,0 +1,30 @@
+import os
+
+
+def get_op_num_threads(env_name: str) -> int:
+    env_value = os.getenv(env_name, None)
+    return get_value_from_string(env_value, -1)
+
+
+def get_load_images_timeout() -> int:
+    env_value = os.getenv('MINERU_PDF_LOAD_IMAGES_TIMEOUT', None)
+    return get_value_from_string(env_value, 300)
+
+
+def get_value_from_string(env_value: str, default_value: int) -> int:
+    if env_value is not None:
+        try:
+            num_threads = int(env_value)
+            if num_threads > 0:
+                return num_threads
+        except ValueError:
+            return default_value
+    return default_value
+
+
+if __name__ == '__main__':
+    print(get_value_from_string('1', -1))
+    print(get_value_from_string('0', -1))
+    print(get_value_from_string('-1', -1))
+    print(get_value_from_string('abc', -1))
+    print(get_load_images_timeout())
\ No newline at end of file
diff --git a/mineru/utils/pdf_image_tools.py b/mineru/utils/pdf_image_tools.py
index 5bbaeadb..f8709926 100644
--- a/mineru/utils/pdf_image_tools.py
+++ b/mineru/utils/pdf_image_tools.py
@@ -9,6 +9,7 @@ from PIL import Image
 
 from mineru.data.data_reader_writer import FileBasedDataWriter
 from mineru.utils.check_sys_env import is_windows_environment
+from mineru.utils.os_env_config import get_load_images_timeout
 from mineru.utils.pdf_reader import image_to_b64str, image_to_bytes, page_to_image
 from mineru.utils.enum_class import ImageType
 from mineru.utils.hash_utils import str_sha256
@@ -51,7 +52,7 @@ def load_images_from_pdf(
         start_page_id=0,
         end_page_id=None,
         image_type=ImageType.PIL,
-        timeout=300,
+        timeout=None,
         threads=4,
 ):
     """带超时控制的 PDF 转图片函数,支持多进程加速
@@ -79,6 +80,8 @@ def load_images_from_pdf(
             image_type
         ), pdf_doc
     else:
+        if timeout is None:
+            timeout = get_load_images_timeout()
         end_page_id = get_end_page_id(end_page_id, len(pdf_doc))
 
         # 计算总页数

From dae2cc8514a1585b0518198749151010aeb6dfd6 Mon Sep 17 00:00:00 2001
From: Xiaomeng Zhao <moe@myhloli.com>
Date: Tue, 4 Nov 2025 19:59:59 +0800
Subject: [PATCH 25/28] Update mineru/utils/pdf_image_tools.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 mineru/utils/pdf_image_tools.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mineru/utils/pdf_image_tools.py b/mineru/utils/pdf_image_tools.py
index f8709926..41d0237f 100644
--- a/mineru/utils/pdf_image_tools.py
+++ b/mineru/utils/pdf_image_tools.py
@@ -63,7 +63,7 @@ def load_images_from_pdf(
         start_page_id (int, optional): 起始页码. Defaults to 0.
         end_page_id (int | None, optional): 结束页码. Defaults to None.
         image_type (ImageType, optional): 图片类型. Defaults to ImageType.PIL.
-        timeout (int): 超时时间(秒),默认 300 秒
+        timeout (int | None, optional): 超时时间(秒)。如果为 None，则从环境变量 MINERU_PDF_LOAD_IMAGES_TIMEOUT 读取，若未设置则默认为 300 秒。
         threads (int): 进程数,默认 4
 
     Raises:

From 5ec07ee7aba67401ca257de4e8cdf5c5bbe2a584 Mon Sep 17 00:00:00 2001
From: myhloli <moe@myhloli.com>
Date: Tue, 4 Nov 2025 20:18:14 +0800
Subject: [PATCH 26/28] feat: update environment variable for PDF rendering
 timeout and enhance documentation

---
 README.md                     | 4 ++++
 README_zh-CN.md               | 4 ++++
 docs/en/usage/cli_tools.md    | 2 +-
 docs/zh/usage/cli_tools.md    | 2 +-
 mineru/utils/os_env_config.py | 2 +-
 5 files changed, 11 insertions(+), 3 deletions(-)

diff --git a/README.md b/README.md
index 6dc6b9a1..83ac47df 100644
--- a/README.md
+++ b/README.md
@@ -44,6 +44,10 @@
 </div>
 
 # Changelog
+- 2025/11/04 2.6.4 Release
+  - Added timeout configuration for PDF image rendering, default is 300 seconds, can be configured via environment variable `MINERU_PDF_RENDER_TIMEOUT` to prevent long blocking of the rendering process caused by some abnormal PDF files.
+  - Added CPU thread count configuration options for ONNX models, default is the system CPU core count, can be configured via environment variables `MINERU_INTRA_OP_NUM_THREADS` and `MINERU_INTER_OP_NUM_THREADS` to reduce CPU resource contention conflicts in high concurrency scenarios.
+
 - 2025/10/31 2.6.3 Release
   - Added support for a new backend `vlm-mlx-engine`, enabling MLX-accelerated inference for the MinerU2.5 model on Apple Silicon devices. Compared to the `vlm-transformers` backend, `vlm-mlx-engine` delivers a 100%–200% speed improvement.
   - Bug fixes: #3849, #3859
diff --git a/README_zh-CN.md b/README_zh-CN.md
index b36f7332..e9dea28a 100644
--- a/README_zh-CN.md
+++ b/README_zh-CN.md
@@ -44,6 +44,10 @@
 </div>
 
 # 更新记录
+- 2025/11/04 2.6.4 发布
+  - 为pdf渲染图片增加超时配置，默认为300秒，可通过环境变量`MINERU_PDF_RENDER_TIMEOUT`进行配置，防止部分异常pdf文件导致渲染过程长时间阻塞。
+  - 为onnx模型增加cpu线程数配置选项，默认为系统cpu核心数，可通过环境变量`MINERU_INTRA_OP_NUM_THREADS`和`MINERU_INTER_OP_NUM_THREADS`进行配置，以减少高并发场景下的对cpu资源的抢占冲突。
+
 - 2025/10/31 2.6.3 发布
   - 增加新后端`vlm-mlx-engine`支持，在Apple Silicon设备上支持使用`MLX`加速`MinerU2.5`模型推理，相比`vlm-transformers`后端，`vlm-mlx-engine`后端速度提升100%~200%。
   - bug修复:  #3849  #3859
diff --git a/docs/en/usage/cli_tools.md b/docs/en/usage/cli_tools.md
index 5b81027e..ccffbc43 100644
--- a/docs/en/usage/cli_tools.md
+++ b/docs/en/usage/cli_tools.md
@@ -100,7 +100,7 @@ Here are the environment variables and their descriptions:
     * Used to enable table merging functionality
     * Default is `true`, can be set to `false` via environment variable to disable table merging functionality.
 
-- `MINERU_PDF_LOAD_IMAGES_TIMEOUT`:
+- `MINERU_PDF_RENDER_TIMEOUT`:
     * Used to set the timeout period (in seconds) for rendering PDF to images
     * Default is `300` seconds, can be set to other values via environment variable to adjust the image rendering timeout.
 
diff --git a/docs/zh/usage/cli_tools.md b/docs/zh/usage/cli_tools.md
index 15d4d2eb..218da1e2 100644
--- a/docs/zh/usage/cli_tools.md
+++ b/docs/zh/usage/cli_tools.md
@@ -95,7 +95,7 @@ MinerU命令行工具的某些参数存在相同功能的环境变量配置，
     * 用于启用表格合并功能
     * 默认为`true`，可通过环境变量设置为`false`来禁用表格合并功能。
 
-- `MINERU_PDF_LOAD_IMAGES_TIMEOUT`：
+- `MINERU_PDF_RENDER_TIMEOUT`：
     * 用于设置将PDF渲染为图片的超时时间（秒）
     * 默认为`300`秒，可通过环境变量设置为其他值以调整渲染图片的超时时间。
 
diff --git a/mineru/utils/os_env_config.py b/mineru/utils/os_env_config.py
index 684976ca..43d01334 100644
--- a/mineru/utils/os_env_config.py
+++ b/mineru/utils/os_env_config.py
@@ -7,7 +7,7 @@ def get_op_num_threads(env_name: str) -> int:
 
 
 def get_load_images_timeout() -> int:
-    env_value = os.getenv('MINERU_PDF_LOAD_IMAGES_TIMEOUT', None)
+    env_value = os.getenv('MINERU_PDF_RENDER_TIMEOUT', None)
     return get_value_from_string(env_value, 300)
 
 

From fe1549960d14951834a748ff721096f5e022f9ba Mon Sep 17 00:00:00 2001
From: Xiaomeng Zhao <moe@myhloli.com>
Date: Tue, 4 Nov 2025 20:20:37 +0800
Subject: [PATCH 27/28] Update mineru/utils/pdf_image_tools.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 mineru/utils/pdf_image_tools.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mineru/utils/pdf_image_tools.py b/mineru/utils/pdf_image_tools.py
index 41d0237f..7347b9d5 100644
--- a/mineru/utils/pdf_image_tools.py
+++ b/mineru/utils/pdf_image_tools.py
@@ -97,7 +97,7 @@ def load_images_from_pdf(
         for i in range(actual_threads):
             range_start = start_page_id + i * pages_per_thread
             if i == actual_threads - 1:
-                # 最后一个线程处理剩余所有页面
+                # 最后一个进程处理剩余所有页面
                 range_end = end_page_id
             else:
                 range_end = start_page_id + (i + 1) * pages_per_thread - 1

From e010b0974ac39dbc32fa8586b754d0eac7c22cf5 Mon Sep 17 00:00:00 2001
From: Xiaomeng Zhao <moe@myhloli.com>
Date: Tue, 4 Nov 2025 20:21:37 +0800
Subject: [PATCH 28/28] Update mineru/utils/pdf_image_tools.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 mineru/utils/pdf_image_tools.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mineru/utils/pdf_image_tools.py b/mineru/utils/pdf_image_tools.py
index 7347b9d5..450f024d 100644
--- a/mineru/utils/pdf_image_tools.py
+++ b/mineru/utils/pdf_image_tools.py
@@ -88,7 +88,7 @@ def load_images_from_pdf(
         total_pages = end_page_id - start_page_id + 1
 
         # 实际使用的进程数不超过总页数
-        actual_threads = min(min(os.cpu_count() or 1, threads), total_pages)
+        actual_threads = min(os.cpu_count() or 1, threads, total_pages)
 
         # 根据实际进程数分组页面范围
         pages_per_thread = max(1, total_pages // actual_threads)