mem: reduce PaddleOCR rec_batch_num from 6 to 1

CHANGELOG.md

@@ -1,3 +1,8 @@
+## 0.22.8
+
+### Enhancements
+- **Reduce PaddleOCR memory usage**: Set `rec_batch_num=1` when creating PaddleOCR instances to reduce Paddle inference engine arena allocation from 4x500 MiB to 1x500 MiB during text recognition
+
 ## 0.22.7
 
 ### Fixes

unstructured/__version__.py

@@ -1 +1 @@
-__version__ = "0.22.7"  # pragma: no cover
+__version__ = "0.22.8"  # pragma: no cover

unstructured/partition/utils/ocr_models/paddle_ocr.py

@@ -48,6 +48,7 @@ def load_agent(self, language: str):
                 lang=language,
                 enable_mkldnn=True,
                 show_log=False,
+                rec_batch_num=1,
             )
         except AttributeError:
             paddle_ocr = PaddleOCR(
@@ -56,6 +57,7 @@ def load_agent(self, language: str):
                 lang=language,
                 enable_mkldnn=False,
                 show_log=False,
+                rec_batch_num=1,
             )
         return paddle_ocr