Merge remote-tracking branch 'origin/main' into feat-multiple-ocr-engines

CHANGELOG.md

@@ -1,3 +1,9 @@
+## [v1.18.0](https://github.com/DS4SD/docling/releases/tag/v1.18.0) - 2024-10-03
+
+### Feature
+
+* New torch-based docling models ([#120](https://github.com/DS4SD/docling/issues/120)) ([`2422f70`](https://github.com/DS4SD/docling/commit/2422f706a1b02a679bcbaaba097fef2f69aba0f4))
+
 ## [v1.17.0](https://github.com/DS4SD/docling/releases/tag/v1.17.0) - 2024-10-03
 
 ### Feature

docling/document_converter.py

@@ -199,9 +199,6 @@ class DocumentConverter:
                 end_pb_time = time.time() - start_pb_time
                 _log.info(f"Finished converting page batch time={end_pb_time:.3f}")
 
-            # Free up mem resources of PDF backend
-            in_doc._backend.unload()
-
             conv_res.pages = all_assembled_pages
             self._assemble_doc(conv_res)
 
@@ -227,6 +224,11 @@ class DocumentConverter:
                 f"{trace}"
             )
 
+        finally:
+            # Always unload the PDF backend, even in case of failure
+            if in_doc._backend:
+                in_doc._backend.unload()
+
         end_doc_time = time.time() - start_doc_time
         _log.info(
             f"Finished converting document time-pages={end_doc_time:.2f}/{in_doc.page_count}"

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "docling"
-version = "1.17.0"  # DO NOT EDIT, updated automatically
+version = "1.18.0"  # DO NOT EDIT, updated automatically
 description = "Docling PDF conversion package"
 authors = ["Christoph Auer <cau@zurich.ibm.com>", "Michele Dolfi <dol@zurich.ibm.com>", "Maxim Lysak <mly@zurich.ibm.com>", "Nikos Livathinos <nli@zurich.ibm.com>", "Ahmed Nassar <ahn@zurich.ibm.com>", "Peter Staar <taa@zurich.ibm.com>"]
 license = "MIT"