Update doc/image (1).png, doc/image (10).png, doc/image (2).png, doc/image...

Update doc/image (1).png, doc/image (10).png, doc/image (2).png, doc/image (3).png, doc/image (4).png, doc/image (5).png, doc/image (6).png, doc/image (7).png, doc/image (8).png, doc/image (9).png, doc/image.png, doc/image11.png, doc/image12.png, magic_pdf/tools/pdf_server.py, magic_pdf/tools/common.py, magic_pdf/pipe/UNIPipe.py, magic_pdf/pipe/AbsPipe.py, magic_pdf/dict2md/ocr_mkcontent.py, magic_pdf/dict2md/ocr_server.py, magic_pdf/config.ini, README.md files
Deleted magic_pdf/tools/pdf_client.py

README.md

@@ -50,7 +50,19 @@
     <img src="doc/image (10).png"/>
 </div>
 
-### 5、启动qwen-ocr模块：
+### 5、配置config.ini中的路由地址
+
+vim magic_pdf/config.ini
+
+默认如下：
+
+`pdf_server = http://0.0.0.0:4090`
+
+`ocr_server = http://0.0.0.0:4080`
+
+根据需要，自行配置路由地址
+
+### 6、启动qwen-ocr模块：
 
 修改magic_pdf/magic_pdf/dict2md/ocr_server.py文件中模型路径地址
 
@@ -62,20 +74,20 @@
 
 `python magic_pdf/dict2md/ocr_server.py`
 
-默认使用6020端口，0号DCU卡 ，可以通过--dcu_id 指定卡，--server_port指定端口号，-c 指定qwen模型地址
+默认使用0号DCU卡 ，可以通过--dcu_id 指定卡，-c 指定qwen模型地址，--config_path 指定config.ini路径
 
 qwen-ocr模块启动成功：
 <div align=center>
     <img src="doc/image (5).png"/>
 </div>
 
-### 5、启动pdf-server解析服务：
+### 7、启动pdf-server解析服务：
 
 #### pdf-server解析服务启动代码：
 
 `python magic_pdf/tools/pdf_server.py`
 
-默认使用6030端口，0号DCU卡 ，可以通过--dcu_id 指定卡，--pdf_port指定端口号
+默认使用0号DCU卡 ，可以通过--dcu_id 指定卡，--config_path 指定config.ini路径
 
 <div align=center>
     <img src="doc/image (6).png"/>
@@ -86,9 +98,18 @@ qwen-ocr模块启动成功：
     <img src="doc/image (7).png"/>
 </div>
 
-### 6、解析pdf
+### 8、解析pdf
+
 `python magic_pdf/parse/common_parse.py -p [文件/目录 路径] -o [输出地址]`
+
+-p指定pdf路径，-o指定输出路径 --config_path 指定config.ini路径 
+
+<div align=center>
+    <img src="doc/image12.png"/>
+</div>
+
 <div align=center>
     <img src="doc/image (8).png"/>
 </div>
--p指定pdf路径，-o指定输出路径
+
+

magic_pdf/config.ini

+[server]
+pdf_server = http://0.0.0.0:4090
+ocr_server = http://0.0.0.0:4080

magic_pdf/dict2md/ocr_mkcontent.py

+import configparser
 import re
 import time
 
@@ -120,11 +121,14 @@ def ocr_mk_markdown_with_para_core(paras_of_layout, mode, img_buket_path=''):
     return page_markdown
 
 
-def ocr_mk_markdown_with_para_core_v2(paras_of_layout,
+def ocr_mk_markdown_with_para_core_v2(config_path,paras_of_layout,
                                       mode,
                                       img_buket_path=''):
     page_markdown = []
-    url = 'http://127.0.0.1:6020'
+    config = configparser.ConfigParser()
+    config.read(config_path)
+    url = config.get('server', 'ocr_server')
+    logger.info(f'ocr_server：{url}')
     client = PredictClient(url)
     for para_block in paras_of_layout:
         para_text = ''
@@ -415,7 +419,8 @@ def ocr_mk_mm_standard_format(pdf_info_dict: list):
     return content_list
 
 
-def union_make(pdf_info_dict: list,
+def union_make(config_path: str,
+               pdf_info_dict: list,
                make_mode: str,
                drop_mode: str,
                img_buket_path: str = ''):
@@ -442,11 +447,11 @@ def union_make(pdf_info_dict: list,
             continue
         if make_mode == MakeMode.MM_MD:
             page_markdown = ocr_mk_markdown_with_para_core_v2(
-                paras_of_layout, 'mm', img_buket_path)
+                config_path,paras_of_layout, 'mm', img_buket_path)
             output_content.extend(page_markdown)
         elif make_mode == MakeMode.NLP_MD:
             page_markdown = ocr_mk_markdown_with_para_core_v2(
-                paras_of_layout, 'nlp')
+                config_path,paras_of_layout, 'nlp')
             output_content.extend(page_markdown)
         elif make_mode == MakeMode.STANDARD_FORMAT:
             for para_block in paras_of_layout:

magic_pdf/dict2md/ocr_server.py

@@ -2,10 +2,12 @@
 #
 # This source code is licensed under the license found in the
 # LICENSE file in the root directory of this source tree
-
+import configparser
 import copy
 import re
 import gc
+import time
+
 import torch
 from argparse import ArgumentParser
 from threading import Thread
@@ -15,6 +17,7 @@ from transformers import AutoProcessor, Qwen2VLForConditionalGeneration, TextIte
 from fastapi import FastAPI
 from pydantic import BaseModel
 from typing import Optional
+from loguru import logger
 
 app = FastAPI()
 
@@ -36,10 +39,11 @@ def _get_args():
                         help='Create a publicly shareable link for the interface.')
     parser.add_argument('--inbrowser', action='store_true', default=False,
                         help='Automatically launch the interface in a new tab on the default browser.')
-    parser.add_argument('--server_port', type=int, default=6020, help='Demo server port.')
-    parser.add_argument('--server_name', type=str, default='127.0.0.1', help='Demo server name.')
     parser.add_argument('--dcu_id', type=str, default='0', help='Specify the GPU ID to load the model onto.')
-
+    parser.add_argument(
+        '--config_path',
+        default='/home/practice/magic_pdf-main/magic_pdf/config.ini',
+        )
     args = parser.parse_args()
     return args
 
@@ -232,13 +236,14 @@ async def predict(item: Item):
             ]
         }
     ]
-
+    start = time.time()
     generated_text = ''
     for response in call_local_model(model, processor, messages):
         generated_text = _parse_text(response)
 
     _gc()
-
+    end = time.time()
+    logger.info(f'【{item.image_path}】解析的结果是：{generated_text},耗时为：{end-start}')
     return {"Generated Text": generated_text}
 
 
@@ -246,6 +251,12 @@ if __name__ == "__main__":
     import uvicorn
 
     args = _get_args()
-    uvicorn.run(app, host=args.server_name, port=args.server_port)
+    config = configparser.ConfigParser()
+    config.read(args.config_path)
+    ocr_server = config.get('server', 'ocr_server')
+    if 'http' in ocr_server:
+        ocr_server = ocr_server.split('://')[1]
+    host,port = ocr_server.split(':')[0],int(ocr_server.split(':')[1])
+    uvicorn.run(app, host=host, port=port)
 
 

magic_pdf/pipe/AbsPipe.py

@@ -56,9 +56,9 @@ class AbsPipe(ABC):
         content_list = AbsPipe.mk_uni_format(self.get_compress_pdf_mid_data(), img_parent_path, drop_mode)
         return content_list
 
-    def pipe_mk_markdown(self, img_parent_path: str, drop_mode=DropMode.WHOLE_PDF, md_make_mode=MakeMode.MM_MD):
+    def pipe_mk_markdown(self,config_path,img_parent_path: str, drop_mode=DropMode.WHOLE_PDF, md_make_mode=MakeMode.MM_MD):
         # logger.info(f'self.get_compress_pdf_mid_data():\n{self.get_compress_pdf_mid_data()}')
-        md_content = AbsPipe.mk_markdown(self.get_compress_pdf_mid_data(), img_parent_path, drop_mode, md_make_mode)
+        md_content = AbsPipe.mk_markdown(config_path,self.get_compress_pdf_mid_data(), img_parent_path, drop_mode, md_make_mode)
         return md_content
 
     @staticmethod
@@ -101,16 +101,13 @@ class AbsPipe(ABC):
         return content_list
 
     @staticmethod
-    def mk_markdown(compressed_pdf_mid_data: str, img_buket_path: str, drop_mode=DropMode.WHOLE_PDF, md_make_mode=MakeMode.MM_MD) -> list:
+    def mk_markdown(config_path: str,compressed_pdf_mid_data: str, img_buket_path: str, drop_mode=DropMode.WHOLE_PDF, md_make_mode=MakeMode.MM_MD) -> list:
         """
         根据pdf类型，markdown
         """
         pdf_mid_data = JsonCompressor.decompress_json(compressed_pdf_mid_data)
         pdf_info_list = pdf_mid_data["pdf_info"]
-        logger.info(f'pdf_mid_data:\n{pdf_mid_data}')
-        logger.info('-'*80)
-        logger.info(f'pdf_info_list:\n{pdf_info_list}')
-        md_content = union_make(pdf_info_list, md_make_mode, drop_mode, img_buket_path)
+        md_content = union_make(config_path,pdf_info_list, md_make_mode, drop_mode, img_buket_path)
         return md_content
 
 

magic_pdf/pipe/UNIPipe.py

@@ -52,11 +52,11 @@ class UNIPipe(AbsPipe):
         logger.info("uni_pipe mk content list finished")
         return result
 
-    def pipe_mk_markdown(self, img_parent_path: str, drop_mode=DropMode.WHOLE_PDF, md_make_mode=MakeMode.MM_MD):
+    def pipe_mk_markdown(self,config_path, img_parent_path: str, drop_mode=DropMode.WHOLE_PDF, md_make_mode=MakeMode.MM_MD):
        # logger.info(f'img_parent_path：\n{img_parent_path}')
        # logger.info(f'drop_mode：\n{drop_mode}')
       #  logger.info(f'md_make_mode：\n{md_make_mode}')
-        result = super().pipe_mk_markdown(img_parent_path, drop_mode, md_make_mode)
+        result = super().pipe_mk_markdown(config_path,img_parent_path, drop_mode, md_make_mode)
         logger.info(f"uni_pipe mk {md_make_mode} finished")
         return result
 

magic_pdf/tools/common.py

@@ -40,12 +40,14 @@ def remove_empty_lines_from_file(file_path):
         file.writelines(non_empty_lines)
 
 def do_parse(
+    config_path,
     output_dir,
     pdf_file_name,
     pdf_bytes,
     model_list,
     parse_method,
     debug_able,
+
     f_draw_span_bbox=True,
     f_draw_layout_bbox=True,
     f_dump_md=True,
@@ -57,6 +59,7 @@ def do_parse(
     f_draw_model_bbox=False,
     start_page_id=0,
     end_page_id=None,
+
 ):
     if debug_able:
         logger.warning('debug mode is on')
@@ -72,7 +75,7 @@ def do_parse(
     image_dir = str(os.path.basename(local_image_dir))
     # logger.info(f'model_list:{model_list}')
     # logger.info(f'local_image_dir:::{local_image_dir}')
-    logger.info(f'image_dir:::{image_dir}')
+    # logger.info(f'image_dir:::{image_dir}')
     if parse_method == 'auto':
         jso_useful_key = {'_pdf_type': '', 'model_list': model_list}
         pipe = UNIPipe(pdf_bytes, jso_useful_key, image_writer, is_debug=True,
@@ -108,7 +111,7 @@ def do_parse(
     # if f_draw_model_bbox:
     #     drow_model_bbox(copy.deepcopy(orig_model_list), pdf_bytes, local_md_dir, pdf_file_name)
 
-    md_content = pipe.pipe_mk_markdown(local_image_dir,
+    md_content = pipe.pipe_mk_markdown(config_path,local_image_dir,
                                        drop_mode=DropMode.NONE,
                                        md_make_mode=f_make_md_mode)
     

magic_pdf/tools/pdf_client.py

-# -*- coding: utf-8 -*-
-import time
-
-import requests
-from loguru import logger
-import argparse
-import os
-
-
-
-class ocrPdfClient:
-    def __init__(self, api_url):
-        self.api_url = api_url
-
-    def ocr_pdf_client(self, path,output_dir):
-        payload = {
-            "path": str(path),
-            "output_dir": str(output_dir),
-        }
-        logger.info(f'pdf路径:{path}，输出路径{output_dir}')
-        response = requests.post(f"{self.api_url}/pdf_ocr", json=payload)
-        logger.info(f'response:{response}')
-        if response.status_code == 200:
-            return output_dir
-
-        else:
-            raise Exception(f"ocrPdf API request failed with status code {response.status_code}")
-
-
-def parse_args():
-    parser = argparse.ArgumentParser()
-    parser.add_argument(
-        '--url',
-        default='http://0.0.0.0:6030',
-        )
-    parser.add_argument(
-        '--path',
-        '-p',
-        required=True
-        )
-    parser.add_argument(
-        '--output_dir',
-        '-o',
-        required=True
-        )
-    args = parser.parse_args()
-    return args
-
-
-def main():
-    args = parse_args()
-
-    embedder = ocrPdfClient(args.url)
-    doc_analyze_start = time.time()
-
-    if not os.path.isabs(args.output_dir):
-        current_working_directory = os.getcwd()
-        output_dir = os.path.join(current_working_directory, args.output_dir)
-        # logger.info(f'相对路径output_dir:{output_dir}')
-    else:
-        output_dir = args.output_dir
-    logger.info(f'output_dir:{output_dir}')
-    try:
-        res = embedder.ocr_pdf_client(path=args.path,output_dir=output_dir)
-        if res:
-            logger.info(f"output_dir: '{res}'")
-        else:
-            logger.warning("None")
-    except requests.exceptions.RequestException as e:
-        logger.error(f"Error while making request to reranker service: {e}")
-    except Exception as e:
-        logger.error(f"Unexpected error occurred: {e}")
-    doc_analyze_cost = time.time() - doc_analyze_start
-
-    logger.info(f'解析当前pdf{args.path}耗时为:{doc_analyze_cost}')
-
-if __name__ == "__main__":
-    main()
-

magic_pdf/tools/pdf_server.py

@@ -14,13 +14,15 @@ from argparse import ArgumentParser
 from pydantic import BaseModel
 import uvicorn
 import time
-
+import configparser
+#from magic_pdf.tools.config import update_config
 
 app = FastAPI()
 method = 'auto'
 class ocrRequest(BaseModel):
     path: str
     output_dir: str
+    config_path: str
 
 def parse_args():
     parser = ArgumentParser()
@@ -28,10 +30,6 @@ def parse_args():
         '--dcu_id',
         default='0',
         help='设置DCU')
-    parser.add_argument(
-        '--pdf_port',
-        default=6030,
-        help='设置DCU')
     parser.add_argument(
         '--method',
         type=parse_pdf_methods,
@@ -42,41 +40,37 @@ def parse_args():
         without method specified, auto will be used by default.""",
         default = 'auto',
         )
-
-    # parser.add_argument(
-    #     '--start',
-    #     type=int,
-    #     help='The starting page for PDF parsing, beginning from 0.',
-    #     default=0,
-    # )
-    # parser.add_argument(
-    #     '--end',
-    #     type=int,
-    #     help='The ending page for PDF parsing, beginning from 0.',
-    #     default=None,
-    # )
-
     parser.add_argument(
         '--debug',
         type=bool,
         help='Enables detailed debugging information during the execution of the CLI commands.',
         default=False,
     )
+    parser.add_argument(
+        '--config_path',
+        default='/home/practice/magic_pdf-main/magic_pdf/config.ini')
 
     args = parser.parse_args()
     return args
 
 def ocr_pdf_serve(args: str):
     os.environ["CUDA_VISIBLE_DEVICES"] = args.dcu_id
-    uvicorn.run(app, host="0.0.0.0", port=args.pdf_port)
+    config = configparser.ConfigParser()
+    config.read(args.config_path)
+    pdf_server = config.get('server', 'pdf_server')
+    if 'http' in pdf_server:
+        pdf_server = pdf_server.split('://')[1]
+    host,port = pdf_server.split(':')[0],int(pdf_server.split(':')[1])
+    uvicorn.run(app, host=host, port=port)
 
 @app.post("/pdf_ocr")
 # def cli(path, output_dir, method, debug_able, start_page_id, end_page_id):
-def pdf_ocr(request: ocrRequest):
+async def pdf_ocr(request: ocrRequest):
     model_config.__use_inside_model__ = True
     model_config.__model_mode__ = 'full'
     output_dir = request.output_dir
     path = request.path
+    #config_path = request.config_path
     os.makedirs(output_dir, exist_ok=True)
     debug_able = False
     start_page_id = 0
@@ -86,11 +80,12 @@ def pdf_ocr(request: ocrRequest):
         disk_rw = DiskReaderWriter(os.path.dirname(path))
         return disk_rw.read(os.path.basename(path), AbsReaderWriter.MODE_BIN)
 
-    def parse_doc(doc_path: str):
+    def parse_doc(doc_path: str, config_path: str):
         try:
             file_name = str(Path(doc_path).stem)
             pdf_data = read_fn(doc_path)
             do_parse(
+                config_path,
                 output_dir,
                 file_name,
                 pdf_data,
@@ -104,21 +99,8 @@ def pdf_ocr(request: ocrRequest):
         except Exception as e:
             logger.exception(e)
 
-    if os.path.isdir(path):
-        for root, dirs, files in os.walk(path):
-            # 查找所有的pdf文件
-            for file in files:
-                if file.endswith('.pdf'):
-                    # 打印pdf文件的完整路径
-                    doc_path = os.path.join(root, file)
-                    start = time.time()
-                    logger.info(f'正在解析：{doc_path}')
-                    parse_doc(doc_path)
-                    end = time.time()
-                    logger.info(f'解析：{doc_path}的耗时为：{end -start}')
-    else:
-        logger.info(f'正在解析：{path}')
-        parse_doc(path)
+    logger.info(f'config_path：{request.config_path}')
+    parse_doc(path,request.config_path)
 
 def main():
     args = parse_args()