From f6577084de4c6ac759108da8797ee228c1b12af6 Mon Sep 17 00:00:00 2001
From: Yoann Schneider <yschneider@teklia.com>
Date: Fri, 13 Sep 2024 11:57:48 +0200
Subject: [PATCH] Support the case where the model can predict NER entities but
doesn't in that case
---
dan/ocr/predict/attention.py | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/dan/ocr/predict/attention.py b/dan/ocr/predict/attention.py
index 07b8c5cc..6d94a4d4 100644
--- a/dan/ocr/predict/attention.py
+++ b/dan/ocr/predict/attention.py
@@ -44,7 +44,8 @@ def compute_offsets_by_level(full_text: str, level: Level, text_list: List[str])
"""
# offsets[idx] = number of characters between text_list[idx-1] and text_list[idx]
offsets = [int(level != Level.Char)] * (len(text_list) - 1)
- if level == Level.NER:
+ # Take entities into account if there are any
+ if level == Level.NER and text_list:
# Start after the first entity
cursor = len(text_list[0])
for idx, split in enumerate(text_list[1:]):
--
GitLab