improve toc

README.md

@@ -12,25 +12,24 @@ This repository contains the official implementation of our [paper](https://arxi
 **"Open-ended VQA benchmarking of Vision-Language models by exploiting Classification datasets and their semantic hierarchy"**
 accepted to ICLR 2024.
 
-<!--TOC-->
-
-- [oVQA: Open-ended Visual Question Answering](#ovqa-open-ended-visual-question-answering)
-  - [News](#news)
-  - [Setup](#setup)
-    - [Python environment](#python-environment)
-    - [Notes](#notes)
-    - [Setup the paths](#setup-the-paths)
-    - [Setup datasets](#setup-datasets)
-    - [Setup model outputs](#setup-model-outputs)
-    - [Setup model checkpoints](#setup-model-checkpoints)
-    - [Setup metrics](#setup-metrics)
-  - [Running](#running)
-  - [Dev commands](#dev-commands)
-  - [Licenses](#licenses)
-  - [Acknowledgements](#acknowledgements)
-  - [Citation](#citation)
-
-<!--TOC-->
+<!--TOC-->
+
+- [News](#news)
+- [Setup](#setup)
+  - [Python environment](#python-environment)
+  - [Notes](#notes)
+  - [Setup the paths](#setup-the-paths)
+  - [Setup datasets](#setup-datasets)
+  - [Setup model outputs](#setup-model-outputs)
+  - [Setup model checkpoints](#setup-model-checkpoints)
+  - [Setup metrics](#setup-metrics)
+- [Running](#running)
+- [Dev commands](#dev-commands)
+- [Licenses](#licenses)
+- [Acknowledgements](#acknowledgements)
+- [Citation](#citation)
+
+<!--TOC-->
 
 
 ## News

update_toc.py

@@ -0,0 +1,40 @@
+"""
+Same logic as md-toc package except skipping h1 and forcing unix newlines.
+"""
+
+from pathlib import Path
+
+import md_toc
+
+
+def main():
+    file = Path("README.md")
+    toc = md_toc.build_toc(file.as_posix(), keep_header_levels=3)
+    content = file.read_text().replace("\r\n", "\n")
+    marker = "<!--TOC-->"
+    marker_splits = content.split(marker)
+    if len(marker_splits) == 2:
+        before, after = marker_splits
+    elif len(marker_splits) == 3:
+        before, _old_toc, after = marker_splits
+    else:
+        raise ValueError(
+            f"Marker {marker} found {len(marker_splits) - 1} times in {file}, expected 1 or 2."
+        )
+    toc = toc.replace("\r\n", "\n")
+    new_toc = []
+    for line in toc.splitlines():
+        if line.startswith("- "):
+            continue
+        elif line.startswith("  "):
+            line = line[2:]
+        new_toc.append(line)
+    toc = "\n".join(new_toc)
+    print(toc)
+    file.write_text(
+        before + marker + "\n\n" + toc + "\n\n" + marker + after, encoding="utf-8", newline="\n"
+    )
+
+
+if __name__ == "__main__":
+    main()