Merge pull request #2314 from myhloli/dev

refactor(table): replace ocr_engine with lang in table model prediction

README.md

@@ -48,6 +48,9 @@ Easier to use: Just grab MinerU Desktop. No coding, no login, just a simple inte
 </div>
 
 # Changelog
+- 2025/04/22 1.3.7 Released
+  - Fixed the issue where the `lang` parameter was ineffective during table parsing model initialization.
+  - Fixed the significant slowdown in OCR and table parsing speed in `cpu` mode.
 - 2025/04/16 1.3.4 Released
   - Slightly improved the speed of OCR detection by removing some unused blocks.
   - Fixed page-level sorting errors caused by footnotes in certain cases.

README_zh-CN.md

@@ -47,6 +47,9 @@
 </div>
 
 # 更新记录
+- 2025/04/22 1.3.7 发布
+  - 修复表格解析模型初始化时lang参数失效的问题
+  - 修复在`cpu`模式下ocr和表格解析速度大幅下降的问题
 - 2025/04/16 1.3.4 发布
   - 通过移除一些无用的块，小幅提升了ocr-det的速度
   - 修复部分情况下由footnote导致的页面内排序错误

magic_pdf/model/batch_analyze.py

@@ -161,20 +161,13 @@ class BatchAnalyze:
             for table_res_dict in tqdm(table_res_list_all_page, desc="Table Predict"):
                 _lang = table_res_dict['lang']
                 atom_model_manager = AtomModelSingleton()
-                ocr_engine = atom_model_manager.get_atom_model(
-                    atom_model_name='ocr',
-                    ocr_show_log=False,
-                    det_db_box_thresh=0.5,
-                    det_db_unclip_ratio=1.6,
-                    lang=_lang
-                )
                 table_model = atom_model_manager.get_atom_model(
                     atom_model_name='table',
                     table_model_name='rapid_table',
                     table_model_path='',
                     table_max_time=400,
                     device='cpu',
-                    ocr_engine=ocr_engine,
+                    lang=_lang,
                     table_sub_model_name='slanet_plus'
                 )
                 html_code, table_cell_bboxes, logic_points, elapse = table_model.predict(table_res_dict['table_img'])

magic_pdf/model/sub_modules/ocr/paddleocr2pytorch/pytorch_paddle.py

@@ -53,6 +53,11 @@ class PytorchPaddleOCR(TextSystem):
         args = parser.parse_args(args)
 
         self.lang = kwargs.get('lang', 'ch')
+
+        device = get_device()
+        if device == 'cpu' and self.lang == 'ch':
+            self.lang = 'ch_lite'
+
         if self.lang in latin_lang:
             self.lang = 'latin'
         elif self.lang in arabic_lang: