mineru 2.7.4__py3-none-any.whl → 2.7.6__py3-none-any.whl

mineru/backend/vlm/utils.py

@@ -24,6 +24,9 @@ def enable_custom_logits_processors() -> bool:
         compute_capability = "8.0"
     elif hasattr(torch, 'mlu') and torch.mlu.is_available():
         compute_capability = "8.0"
+    elif hasattr(torch, 'sdaa') and torch.sdaa.is_available():
+        compute_capability = "8.0"
+    
     else:
         logger.info("CUDA not available, disabling custom_logits_processors")
         return False
@@ -102,4 +105,128 @@ def set_default_batch_size() -> int:
     except Exception as e:
         logger.warning(f'Error determining VRAM: {e}, using default batch_ratio: 1')
         batch_size = 1
-    return batch_size
+    return batch_size
+
+
+def _get_device_config(device_type: str) -> dict | None:
+    """获取不同设备类型的配置参数"""
+
+    # 各设备类型的配置定义
+    DEVICE_CONFIGS = {
+        # "musa": {
+        #     "compilation_config_dict": {
+        #         "cudagraph_capture_sizes": [1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 14, 16, 18, 20, 24, 28, 30],
+        #         "simple_cuda_graph": True
+        #     },
+        #     "block_size": 32,
+        # },
+        "corex": {
+            "compilation_config_dict": {
+                "cudagraph_mode": "FULL_DECODE_ONLY",
+                "level": 0
+            },
+        },
+        "kxpu": {
+            "compilation_config_dict": {
+                "splitting_ops": [
+                    "vllm.unified_attention", "vllm.unified_attention_with_output",
+                    "vllm.unified_attention_with_output_kunlun", "vllm.mamba_mixer2",
+                    "vllm.mamba_mixer", "vllm.short_conv", "vllm.linear_attention",
+                    "vllm.plamo2_mamba_mixer", "vllm.gdn_attention", "vllm.sparse_attn_indexer"
+                ]
+            },
+            "block_size": 128,
+            "dtype": "float16",
+            "distributed_executor_backend": "mp",
+            "enable_chunked_prefill": False,
+            "enable_prefix_caching": False,
+        },
+    }
+
+    return DEVICE_CONFIGS.get(device_type.lower())
+
+
+def _check_server_arg_exists(args: list, arg_name: str) -> bool:
+    """检查命令行参数列表中是否已存在指定参数"""
+    return any(arg == f"--{arg_name}" or arg.startswith(f"--{arg_name}=") for arg in args)
+
+
+def _add_server_arg_if_missing(args: list, arg_name: str, value: str) -> None:
+    """如果参数不存在，则添加到命令行参数列表"""
+    if not _check_server_arg_exists(args, arg_name):
+        args.extend([f"--{arg_name}", value])
+
+
+def _add_server_flag_if_missing(args: list, flag_name: str) -> None:
+    """如果 flag 不存在，则添加到命令行参数列表"""
+    if not _check_server_arg_exists(args, flag_name):
+        args.append(f"--{flag_name}")
+
+
+def _add_engine_kwarg_if_missing(kwargs: dict, key: str, value) -> None:
+    """如果参数不存在，则添加到 kwargs 字典"""
+    if key not in kwargs:
+        kwargs[key] = value
+
+
+def mod_kwargs_by_device_type(kwargs_or_args: dict | list, vllm_mode: str) -> dict | list:
+    """根据设备类型修改 vllm 配置参数
+
+    Args:
+        kwargs_or_args: 配置参数，server 模式为 list，engine 模式为 dict
+        vllm_mode: vllm 运行模式 ("server", "sync_engine", "async_engine")
+
+    Returns:
+        修改后的配置参数
+    """
+    device_type = os.getenv("MINERU_VLLM_DEVICE", "")
+    config = _get_device_config(device_type)
+
+    if config is None:
+        return kwargs_or_args
+
+    if vllm_mode == "server":
+        _apply_server_config(kwargs_or_args, config)
+    else:
+        _apply_engine_config(kwargs_or_args, config, vllm_mode)
+
+    return kwargs_or_args
+
+
+def _apply_server_config(args: list, config: dict) -> None:
+    """应用 server 模式的配置"""
+    import json
+
+    for key, value in config.items():
+        if key == "compilation_config_dict":
+            _add_server_arg_if_missing(
+                args, "compilation-config",
+                json.dumps(value, separators=(',', ':'))
+            )
+        else:
+            # 转换 key 格式: block_size -> block-size
+            arg_name = key.replace("_", "-")
+            if arg_name in {"enable-chunked-prefill", "enable-prefix-caching"} and value is False:
+                _add_server_flag_if_missing(args, f"no-{arg_name}")
+                continue
+            _add_server_arg_if_missing(args, arg_name, str(value))
+
+
+def _apply_engine_config(kwargs: dict, config: dict, vllm_mode: str) -> None:
+    """应用 engine 模式的配置"""
+    try:
+        from vllm.config import CompilationConfig
+    except ImportError:
+        raise ImportError("Please install vllm to use the vllm-async-engine backend.")
+
+    for key, value in config.items():
+        if key == "compilation_config_dict":
+            if vllm_mode == "sync_engine":
+                compilation_config = value
+            elif vllm_mode == "async_engine":
+                compilation_config = CompilationConfig(**value)
+            else:
+                continue
+            _add_engine_kwarg_if_missing(kwargs, "compilation_config", compilation_config)
+        else:
+            _add_engine_kwarg_if_missing(kwargs, key, value)

mineru/backend/vlm/vlm_analyze.py

@@ -6,7 +6,7 @@ import json
 from loguru import logger
 
 from .utils import enable_custom_logits_processors, set_default_gpu_memory_utilization, set_default_batch_size, \
-    set_lmdeploy_backend
+    set_lmdeploy_backend, mod_kwargs_by_device_type
 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
@@ -101,27 +101,7 @@ class ModelSingleton:
                     except ImportError:
                         raise ImportError("Please install vllm to use the vllm-engine backend.")
 
-                    # musa vllm v1 引擎特殊配置
-                    # device = get_device()
-                    # if device_type.startswith("musa"):
-                    #     import torch
-                    #     if torch.musa.is_available():
-                    #         compilation_config = {
-                    #             "cudagraph_capture_sizes": [1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 14, 16, 18, 20, 24, 28, 30],
-                    #             "simple_cuda_graph": True
-                    #         }
-                    #         block_size = 32
-                    #         kwargs["compilation_config"] = compilation_config
-                    #         kwargs["block_size"] = block_size
-
-                    # corex vllm v1 引擎特殊配置
-                    device_type = os.getenv("MINERU_LMDEPLOY_DEVICE", "")
-                    if device_type.lower() == "corex":
-                        compilation_config = {
-                            "cudagraph_mode": "FULL_DECODE_ONLY",
-                            "level": 0
-                        }
-                        kwargs["compilation_config"] = compilation_config
+                    kwargs = mod_kwargs_by_device_type(kwargs, vllm_mode="sync_engine")
 
                     if "compilation_config" in kwargs:
                         if isinstance(kwargs["compilation_config"], str):
@@ -148,28 +128,7 @@ class ModelSingleton:
                     except ImportError:
                         raise ImportError("Please install vllm to use the vllm-async-engine backend.")
 
-
-                    # musa vllm v1 引擎特殊配置
-                    # device = get_device()
-                    # if device.startswith("musa"):
-                    #     import torch
-                    #     if torch.musa.is_available():
-                    #         compilation_config = CompilationConfig(
-                    #             cudagraph_capture_sizes=[1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 14, 16, 18, 20, 24, 28, 30],
-                    #             simple_cuda_graph=True
-                    #         )
-                    #         block_size = 32
-                    #         kwargs["compilation_config"] = compilation_config
-                    #         kwargs["block_size"] = block_size
-
-                    # corex vllm v1 引擎特殊配置
-                    device_type = os.getenv("MINERU_LMDEPLOY_DEVICE", "")
-                    if device_type.lower() == "corex":
-                        compilation_config = CompilationConfig(
-                            cudagraph_mode="FULL_DECODE_ONLY",
-                            level=0
-                        )
-                        kwargs["compilation_config"] = compilation_config
+                    kwargs = mod_kwargs_by_device_type(kwargs, vllm_mode="async_engine")
 
                     if "compilation_config" in kwargs:
                         if isinstance(kwargs["compilation_config"], dict):

mineru/model/mfr/pp_formulanet_plus_m/predict_formula.py

@@ -89,7 +89,11 @@ class FormulaRecognizer(BaseOCRV20):
         return rec_formula
 
     def batch_predict(
-        self, images_mfd_res: list, images: list, batch_size: int = 64
+        self,
+        images_mfd_res: list,
+        images: list,
+        batch_size: int = 64,
+        interline_enable: bool = True,
     ) -> list:
         images_formula_list = []
         mf_image_list = []
@@ -105,6 +109,8 @@ class FormulaRecognizer(BaseOCRV20):
             for idx, (xyxy, conf, cla) in enumerate(
                 zip(mfd_res.boxes.xyxy, mfd_res.boxes.conf, mfd_res.boxes.cls)
             ):
+                if not interline_enable and cla.item() == 1:
+                    continue  # Skip interline regions if not enabled
                 xmin, ymin, xmax, ymax = [int(p.item()) for p in xyxy]
                 new_item = {
                     "category_id": 13 + int(cla.item()),

mineru/model/vlm/vllm_server.py

@@ -1,8 +1,8 @@
 import os
 import sys
 
-from mineru.backend.vlm.utils import set_default_gpu_memory_utilization, enable_custom_logits_processors
-from mineru.utils.config_reader import get_device
+from mineru.backend.vlm.utils import set_default_gpu_memory_utilization, enable_custom_logits_processors, \
+    mod_kwargs_by_device_type
 from mineru.utils.models_download_utils import auto_download_and_get_model_root_path
 
 from vllm.entrypoints.cli.main import main as vllm_main
@@ -14,8 +14,6 @@ def main():
     has_port_arg = False
     has_gpu_memory_utilization_arg = False
     has_logits_processors_arg = False
-    has_block_size_arg = False
-    has_compilation_config = False
     model_path = None
     model_arg_indices = []
 
@@ -27,10 +25,6 @@ def main():
             has_gpu_memory_utilization_arg = True
         if arg == "--logits-processors" or arg.startswith("--logits-processors="):
             has_logits_processors_arg = True
-        if arg == "--block-size" or arg.startswith("--block-size="):
-            has_block_size_arg = True
-        if arg == "--compilation-config" or arg.startswith("--compilation-config="):
-            has_compilation_config = True
         if arg == "--model":
             if i + 1 < len(args):
                 model_path = args[i + 1]
@@ -57,21 +51,7 @@ def main():
     if (not has_logits_processors_arg) and custom_logits_processors:
         args.extend(["--logits-processors", "mineru_vl_utils:MinerULogitsProcessor"])
 
-    # musa vllm v1 引擎特殊配置 
-    # device = get_device()
-    # if device.startswith("musa"):
-    #     import torch
-    #     if torch.musa.is_available():
-    #         if not has_block_size_arg:
-    #             args.extend(["--block-size", "32"])
-    #         if not has_compilation_config:
-    #             args.extend(["--compilation-config", '{"cudagraph_capture_sizes": [1,2,3,4,5,6,7,8,10,12,14,16,18,20,24,28,30], "simple_cuda_graph": true}'])
-
-    # corex vllm v1 引擎特殊配置
-    device_type = os.getenv("MINERU_LMDEPLOY_DEVICE", "")
-    if device_type.lower() == "corex":
-        if not has_compilation_config:
-            args.extend(["--compilation-config", '{"cudagraph_mode": "FULL_DECODE_ONLY", "level": 0}'])
+    args = mod_kwargs_by_device_type(args, vllm_mode="server")
 
     # 重构参数，将模型路径作为位置参数
     sys.argv = [sys.argv[0]] + ["serve", model_path] + args

mineru/utils/block_sort.py

@@ -202,6 +202,10 @@ def model_init(model_name: str):
         if hasattr(torch, 'mlu') and torch.mlu.is_available():
             if torch.mlu.is_bf16_supported():
                 bf_16_support = True
+    elif device_name.startswith("sdaa"):
+        if hasattr(torch, 'sdaa') and torch.sdaa.is_available():
+            if torch.sdaa.is_bf16_supported():
+                bf_16_support = True  
 
     if model_name == 'layoutreader':
         # 检测modelscope的缓存目录是否存在

mineru/utils/config_reader.py

@@ -98,7 +98,12 @@ def get_device():
                             if torch.mlu.is_available():
                                 return "mlu"
                         except Exception as e:
-                            pass
+                            try:
+                                if torch.sdaa.is_available():
+                                    return "sdaa"
+                            except Exception as e:
+                                pass
+                                                           
         return "cpu"
 
 

mineru/utils/model_utils.py

@@ -432,6 +432,9 @@ def clean_memory(device='cuda'):
     elif str(device).startswith("mlu"):
         if torch.mlu.is_available():
             torch.mlu.empty_cache()
+    elif str(device).startswith("sdaa"):
+        if torch.sdaa.is_available():
+            torch.sdaa.empty_cache()  
     gc.collect()
 
 
@@ -476,5 +479,8 @@ def get_vram(device) -> int:
     elif str(device).startswith("mlu"):
         if torch.mlu.is_available():
             total_memory = round(torch.mlu.get_device_properties(device).total_memory / (1024 ** 3))  # 转为 GB
+    elif str(device).startswith("sdaa"):
+        if torch.sdaa.is_available():
+            total_memory = round(torch.sdaa.get_device_properties(device).total_memory / (1024 ** 3))  # 转为 GB          
 
     return total_memory

mineru/utils/os_env_config.py

@@ -11,6 +11,11 @@ def get_load_images_timeout() -> int:
     return get_value_from_string(env_value, 300)
 
 
+def get_load_images_threads() -> int:
+    env_value = os.getenv('MINERU_PDF_RENDER_THREADS', None)
+    return get_value_from_string(env_value, 4)
+
+
 def get_value_from_string(env_value: str, default_value: int) -> int:
     if env_value is not None:
         try:

mineru/utils/pdf_image_tools.py

@@ -1,5 +1,7 @@
 # Copyright (c) Opendatalab. All rights reserved.
 import os
+import signal
+import time
 from io import BytesIO
 
 import numpy as np
@@ -9,13 +11,13 @@ from PIL import Image, ImageOps
 
 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.os_env_config import get_load_images_timeout, get_load_images_threads
 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 concurrent.futures import ProcessPoolExecutor, TimeoutError as FuturesTimeoutError
+from concurrent.futures import ProcessPoolExecutor, wait, ALL_COMPLETED
 
 
 def pdf_page_to_image(page: pdfium.PdfPage, dpi=200, image_type=ImageType.PIL) -> dict:
@@ -57,7 +59,7 @@ def load_images_from_pdf(
     end_page_id=None,
     image_type=ImageType.PIL,
     timeout=None,
-    threads=4,
+    threads=None,
 ):
     """带超时控制的 PDF 转图片函数,支持多进程加速
 
@@ -67,8 +69,8 @@ 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 | 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
+        timeout (int | None, optional): 超时时间(秒)。如果为 None，则从环境变量 MINERU_PDF_RENDER_TIMEOUT 读取，若未设置则默认为 300 秒。
+        threads (int): 进程数, 如果为 None，则从环境变量 MINERU_PDF_RENDER_THREADS 读取，若未设置则默认为 4.
 
     Raises:
         TimeoutError: 当转换超时时抛出
@@ -86,6 +88,9 @@ def load_images_from_pdf(
     else:
         if timeout is None:
             timeout = get_load_images_timeout()
+        if threads is None:
+            threads = get_load_images_threads()
+
         end_page_id = get_end_page_id(end_page_id, len(pdf_doc))
 
         # 计算总页数
@@ -108,11 +113,13 @@ def load_images_from_pdf(
 
             page_ranges.append((range_start, range_end))
 
-        # logger.debug(f"PDF to images using {actual_threads} processes, page ranges: {page_ranges}")
+        logger.debug(f"PDF to images using {actual_threads} processes, page ranges: {page_ranges}")
 
-        with ProcessPoolExecutor(max_workers=actual_threads) as executor:
+        executor = ProcessPoolExecutor(max_workers=actual_threads)
+        try:
             # 提交所有任务
             futures = []
+            future_to_range = {}
             for range_start, range_end in page_ranges:
                 future = executor.submit(
                     _load_images_from_pdf_worker,
@@ -122,27 +129,68 @@ def load_images_from_pdf(
                     range_end,
                     image_type,
                 )
-                futures.append((range_start, future))
-
-            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))
-
-                # 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:
+                futures.append(future)
+                future_to_range[future] = range_start
+
+            # 使用 wait() 设置单一全局超时
+            done, not_done = wait(futures, timeout=timeout, return_when=ALL_COMPLETED)
+
+            # 检查是否有未完成的任务（超时情况）
+            if not_done:
+                # 超时：强制终止所有子进程
+                _terminate_executor_processes(executor)
                 pdf_doc.close()
-                executor.shutdown(wait=False, cancel_futures=True)
                 raise TimeoutError(f"PDF to images conversion timeout after {timeout}s")
 
+            # 所有任务完成，收集结果
+            all_results = []
+            for future in futures:
+                range_start = future_to_range[future]
+                # 这里不需要 timeout，因为任务已完成
+                images_list = future.result()
+                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 Exception as e:
+            # 发生任何异常时，确保清理子进程
+            _terminate_executor_processes(executor)
+            pdf_doc.close()
+            if isinstance(e, TimeoutError):
+                raise
+            raise
+        finally:
+            executor.shutdown(wait=False, cancel_futures=True)
+
+
+def _terminate_executor_processes(executor):
+    """强制终止 ProcessPoolExecutor 中的所有子进程"""
+    if hasattr(executor, '_processes'):
+        for pid, process in executor._processes.items():
+            if process.is_alive():
+                try:
+                    # 先发送 SIGTERM 允许优雅退出
+                    os.kill(pid, signal.SIGTERM)
+                except (ProcessLookupError, OSError):
+                    pass
+
+        # 给子进程一点时间响应 SIGTERM
+        time.sleep(0.1)
+
+        # 对仍然存活的进程发送 SIGKILL 强制终止
+        for pid, process in executor._processes.items():
+            if process.is_alive():
+                try:
+                    os.kill(pid, signal.SIGKILL)
+                except (ProcessLookupError, OSError):
+                    pass
+
 
 def load_images_from_pdf_core(
     pdf_bytes: bytes,

mineru/version.py

@@ -1 +1 @@
-__version__ = "2.7.4"
+__version__ = "2.7.6"

{mineru-2.7.4.dist-info → mineru-2.7.6.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mineru
-Version: 2.7.4
+Version: 2.7.6
 Summary: A practical tool for converting PDF to Markdown
 License: AGPL-3.0
 Project-URL: homepage, https://mineru.net/
@@ -135,17 +135,22 @@ Dynamic: license-file
 
 # Changelog
 
-- 2026/01/30 2.7.4 Release
-  - Added support for domestic computing platforms IluvatarCorex and Cambricon. Currently, the officially supported domestic computing platforms include:
-    - [Ascend](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Ascend/) 
-    - [T-Head](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/THead/) 
-    - [METAX](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/METAX/) 
+- 2026/02/06 2.7.6 Release
+  - Added support for the domestic computing platforms Kunlunxin and Tecorigin; currently, the domestic computing platforms that have been adapted and supported by the official team and vendors include:
+    - [Ascend](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Ascend) 
+    - [T-Head](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/THead) 
+    - [METAX](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/METAX) 
     - [Hygon](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Hygon/)
     - [Enflame](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Enflame/)
     - [MooreThreads](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/MooreThreads/)
     - [IluvatarCorex](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/IluvatarCorex/)
     - [Cambricon](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Cambricon/)
-  - MinerU continues to ensure compatibility with domestic hardware platforms, supporting mainstream chip architectures. With secure and reliable technology, we empower researchers, government, and enterprises to reach new heights in document digitization!
+    - [Kunlunxin](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Kunlunxin/)
+    - [Tecorigin](https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Tecorigin/)  
+  - MinerU continues to support domestic hardware platforms and mainstream chip architectures. With secure and reliable technology, it helps research, government, and enterprise users reach new heights in document digitization!
+
+- 2026/01/30 2.7.4 Release
+  - Added support for domestic computing platforms IluvatarCorex and Cambricon.
 
 - 2026/01/23 2.7.2 Release
   - Added support for domestic computing platforms Hygon, Enflame, and Moore Threads.

{mineru-2.7.4.dist-info → mineru-2.7.6.dist-info}/RECORD

@@ -1,5 +1,5 @@
 mineru/__init__.py,sha256=8CRrCQVuExa0BttRFh3Z40lFy2K5jN0sp67KWjOlj5c,50
-mineru/version.py,sha256=yLdxKZXyzrDqew_33G4dvZoqgGxRCyEx9vhYW3y2Je4,22
+mineru/version.py,sha256=6xG2XfctNZV_iMAbDf3PscewWwjPfwfmAC2zaeMR2KI,22
 mineru/backend/__init__.py,sha256=8CRrCQVuExa0BttRFh3Z40lFy2K5jN0sp67KWjOlj5c,50
 mineru/backend/utils.py,sha256=GLJU3IznDmhE1_qNmkU1UOtsuskIHBezgsEVO6Uar-Y,698
 mineru/backend/hybrid/__init__.py,sha256=IFgr2C8NfSAj8q7JF7QOqMvCiJ6Fc8TIuU3Uh2DaFZU,51
@@ -17,8 +17,8 @@ mineru/backend/pipeline/pipeline_magic_model.py,sha256=w8jGx8f6yZN0Wf2yPP3L9rYKc
 mineru/backend/pipeline/pipeline_middle_json_mkcontent.py,sha256=NJCLGKE7BqM24bRdpXCfTalyiqozowFZjpdzpIUy5aA,14672
 mineru/backend/vlm/__init__.py,sha256=8CRrCQVuExa0BttRFh3Z40lFy2K5jN0sp67KWjOlj5c,50
 mineru/backend/vlm/model_output_to_middle_json.py,sha256=AqYX44gS9crUO_t7SuUatD71EVjow6pI6yA2Ik3gQ0s,5139
-mineru/backend/vlm/utils.py,sha256=PIYqOStLCZlxU9TiZK4EhP90rPYIe_0thEZeP01YPls,3940
-mineru/backend/vlm/vlm_analyze.py,sha256=_2-xJC2C2rT87lZw8JZfC6PFFY0FfEbM9PK2TOkIJao,15604
+mineru/backend/vlm/utils.py,sha256=igxgc-ZXje-TKQvZ2p_YJZTMkHS9yXE7u1-FcaGEVZ0,8523
+mineru/backend/vlm/vlm_analyze.py,sha256=Vc8rRzvcE5egjW_J7L0bueo2dLK3b3KKIzvCK2AyBRk,13500
 mineru/backend/vlm/vlm_magic_model.py,sha256=RodoVwNJhzjyuRLn5Io5gFMIX1NxCuuLzCbUxGaKV80,21447
 mineru/backend/vlm/vlm_middle_json_mkcontent.py,sha256=w-Szbm4HitR7MY4pinSCZZdXtPSqmtlU9cjNh4IOQyg,29499
 mineru/cli/__init__.py,sha256=8CRrCQVuExa0BttRFh3Z40lFy2K5jN0sp67KWjOlj5c,50
@@ -51,7 +51,7 @@ mineru/model/mfd/yolo_v8.py,sha256=OI5AxVgt3FvXp4NYk0BDXXvpDlo9YjM6byDyC_TZ8Js,3
 mineru/model/mfr/__init__.py,sha256=8CRrCQVuExa0BttRFh3Z40lFy2K5jN0sp67KWjOlj5c,50
 mineru/model/mfr/utils.py,sha256=pAi1HnkTuO0R6251Hdl-o50m0wH0Ce89PAf74WCsXPU,11499
 mineru/model/mfr/pp_formulanet_plus_m/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-mineru/model/mfr/pp_formulanet_plus_m/predict_formula.py,sha256=alGX_sPJxZh_7v1sOK3DJ8akfkWO-2c5I_JR7aXMTLU,5588
+mineru/model/mfr/pp_formulanet_plus_m/predict_formula.py,sha256=tYbxdG_oNLb18CsQkusZA-r3fxHQd1uDnfzIFQ6IIU4,5783
 mineru/model/mfr/pp_formulanet_plus_m/processors.py,sha256=MSKyanxiDDjgDQHBov-GjKtPnMx9tSmxBC9GIkM3ft8,23832
 mineru/model/mfr/unimernet/Unimernet.py,sha256=ZK0M9fPmZziK4D33H3YND7RnHiQkRVCS-lvNfY-N7do,7912
 mineru/model/mfr/unimernet/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -151,17 +151,17 @@ mineru/model/utils/tools/infer/predict_system.py,sha256=hkegkn6hq2v2zqHVAP615-k-
 mineru/model/utils/tools/infer/pytorchocr_utility.py,sha256=i1PFN-_kefJUUZ4Vk7igs1TU8gfErTDlDXY6-8Uaurw,9323
 mineru/model/vlm/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 mineru/model/vlm/lmdeploy_server.py,sha256=PvxJNcUIKB8VzWMDXeV1t0SHSgz_ULO36ZAzJbppz90,3262
-mineru/model/vlm/vllm_server.py,sha256=7taySlBANFBTS81Q8PJ6gJWjTgjnFQoGvMknK4NuyLY,3360
+mineru/model/vlm/vllm_server.py,sha256=gC4bkwBbnQXpmxaiq1nPf7RgWF-pUYESjLssveJq6Do,2360
 mineru/resources/header.html,sha256=7xrf6bGloR-3ZeTDyA-JvavE_NeRuUDe3p07cEKUXSI,4769
 mineru/resources/fasttext-langdetect/lid.176.ftz,sha256=jzRyz-hzintgmejpmcPL-uDc0VaWqsfXc4qAOdtgPoM,938013
 mineru/utils/__init__.py,sha256=8CRrCQVuExa0BttRFh3Z40lFy2K5jN0sp67KWjOlj5c,50
 mineru/utils/block_pre_proc.py,sha256=uGBmxf2MR9bplTnQI8xHjCI-kj3plRhJr0hcWKidbOQ,9632
-mineru/utils/block_sort.py,sha256=e6nNjdUeRixT70OfvlEzM1FGwKxFSVwiLtwYGrsG_U0,13724
+mineru/utils/block_sort.py,sha256=5S1VdpRgI72D2dRb3Qp5XQiqSmiPpELwFIqbpshH1jA,13916
 mineru/utils/boxbase.py,sha256=xnGA1k7hVtTQrreqlJmK-SA3y9edTHgLmGiqGrSXckE,7568
 mineru/utils/char_utils.py,sha256=74T5Ylr5mi1uddAIuJku9Z6sH7vhR7t595_H7qmbu4c,1777
 mineru/utils/check_sys_env.py,sha256=TRjzg4xWyoSGrgv4KaP225A-99xBgLAfZ1cPcGqrBAA,1191
 mineru/utils/cli_parser.py,sha256=4seFAu1kulsYnw6WM2q_cxgEOt2tErZVkI-LNEF_kGw,1445
-mineru/utils/config_reader.py,sha256=mwXYVuj52mA__2BU2qOPP0Pn9m0dDLi4mAqPS9a4Pjo,4575
+mineru/utils/config_reader.py,sha256=03ASqJUJIl6CkXVcsewpnPDAo9I7WYdj_hx-osUKrlE,4835
 mineru/utils/cut_image.py,sha256=g3m4nfcJNWlxi-P0kpXTtlmspXkMcLCfGwmYuQ-Z2hE,751
 mineru/utils/draw_bbox.py,sha256=FkgppjUzRhN-uxvChdkhHXcDavJEaApMD6qC6qoRwfQ,20292
 mineru/utils/engine_utils.py,sha256=Jmao9-O-sZDzH7vANKEDaY6NJ8tuthKsTr23LFIeBLU,2203
@@ -172,12 +172,12 @@ mineru/utils/hash_utils.py,sha256=UPS_8NRBmVumdyOv16Lmv6Ly2xK8OVDJEe5gG6gKIFk,85
 mineru/utils/language.py,sha256=7RT3mxSa7jdpoC5ySd7ZddHA7TO7UsnmDOWiYZAxuyg,1433
 mineru/utils/llm_aided.py,sha256=9WUytvxenSAuaWR4sTQhVPQ5h8pY0wVOH1O2sj_6dLs,5149
 mineru/utils/magic_model_utils.py,sha256=8Hv-BDk9Ez4TUx6hrVJ_675yZZggPj6Uib81lSpm0ig,11683
-mineru/utils/model_utils.py,sha256=xlw5hUYKa6o1NiM8PoXO1HFvHfrgY5e4Ut_upGEY9yI,19909
+mineru/utils/model_utils.py,sha256=YadxNuRvuWZ5yW2NkSpD0ZYTJdj0ZVS2X8KF_hlGWCA,20231
 mineru/utils/models_download_utils.py,sha256=UfjvwhxO6BkJHa5JSpEVNZ71GoLMPMmJpym3THET2T4,2957
 mineru/utils/ocr_utils.py,sha256=lPIrwNUib5mrzUkponRYHuUCdjV2qvETNLSzOLyflrU,15990
-mineru/utils/os_env_config.py,sha256=ZNtkR4KrJW72CeIoTNzGDL6tMKv_hL8nzvWIssGWbqY,842
+mineru/utils/os_env_config.py,sha256=VHK9lS3QFJhrwWa9FOFU1Swm7oXnby4SaNNjTyonTTg,990
 mineru/utils/pdf_classify.py,sha256=6DF5pH_9Uq83fsFtp7n4i-OdYQGzoNOV9L0VBUhgBMQ,8078
-mineru/utils/pdf_image_tools.py,sha256=L2kHKoFaQo4CGjS1d68JACrlBycx6gyCnnFlbBFRKuw,8273
+mineru/utils/pdf_image_tools.py,sha256=tTSk39fgJKLEshwPAuJGLl_pVSrmEKjWA55F6dGcr4g,9987
 mineru/utils/pdf_page_id.py,sha256=em966k12CRW4Rj49RGiLB_8ILwkXPBnWRetApax3eTs,400
 mineru/utils/pdf_reader.py,sha256=WeINm5SyWBUXT0wP9lzIbeHs8P6WUIkN6nVL5X4LzG4,3267
 mineru/utils/pdf_text_tool.py,sha256=KEztjfdqsIHHuiTEAMAL7Lr1OS3R7Ur-uTqGiCRjReQ,1364
@@ -185,9 +185,9 @@ mineru/utils/run_async.py,sha256=rPeP4BCZerR8VByRDhiYzfZiahLVqoZEBVAS54dAjNg,128
 mineru/utils/span_block_fix.py,sha256=0eVQjJCrT03woRt9hoh6Uu42Tp1dacfGTv2x3B9qq94,8797
 mineru/utils/span_pre_proc.py,sha256=nu6Bh5TWPKFzHuFfbEs0Asr04M4xOL5IONz_8GJHn44,13862
 mineru/utils/table_merge.py,sha256=LORxz0THemCqH746FMViqEuLzM088M4HgIkEuwDIfNU,21393
-mineru-2.7.4.dist-info/licenses/LICENSE.md,sha256=jVa0BUaKrRH4erV2P5AeJ24I2WRv9chIGxditreJ6e0,34524
-mineru-2.7.4.dist-info/METADATA,sha256=lNxDREB_s7eDnknMUeBn5FCgtDc8qPQS-hEs4Wb6WTg,36928
-mineru-2.7.4.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
-mineru-2.7.4.dist-info/entry_points.txt,sha256=a9AHBIiYe3dpT3oofVQJC8fI0WjDhQASCUlhdMOK120,376
-mineru-2.7.4.dist-info/top_level.txt,sha256=zuGQfZcbsHv4I4oKI9gaKPqEWBFm6xJroKuug2LnKP8,7
-mineru-2.7.4.dist-info/RECORD,,
+mineru-2.7.6.dist-info/licenses/LICENSE.md,sha256=jVa0BUaKrRH4erV2P5AeJ24I2WRv9chIGxditreJ6e0,34524
+mineru-2.7.6.dist-info/METADATA,sha256=m6EbuSPR6iPDZp-fBf90urMSPi9JbGLKZZC5EneGsKc,37245
+mineru-2.7.6.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
+mineru-2.7.6.dist-info/entry_points.txt,sha256=a9AHBIiYe3dpT3oofVQJC8fI0WjDhQASCUlhdMOK120,376
+mineru-2.7.6.dist-info/top_level.txt,sha256=zuGQfZcbsHv4I4oKI9gaKPqEWBFm6xJroKuug2LnKP8,7
+mineru-2.7.6.dist-info/RECORD,,