fix eos bounds when creating paragraph subsets

src/augmenty/doc/subset.py

@@ -51,7 +51,7 @@ def paragraph_subset_augmenter_v1(
     # Respect entity spans
     while start != 0 and example.y[start].ent_iob_ not in {"O", "B", ""}:
         start -= 1
-    while end < doc_len - 1 and example.y[end + 1].ent_iob_ not in {"O", "B", ""}:
+    while end < doc_len - 1 and example.y[end - 1].ent_iob_ not in {"O", "B", ""}:
         end += 1
 
     for k in token_anno: