Fix: img embeddings prompt position

examples/{auto_a_group_of_cyclists_riding_nea_20250711_210229 → auto_a_group_of_cyclists_riding_nea_20250724_104818}/heatmaps.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6f745bd0249ce4c8fde5b105ea3a92acee4b7a915b273f8f38d01d2721fa32ca
-size 11115544
+oid sha256:da1ed693540d83d180048157ca1af810a29128a48c80bfff01494aa114a2036a
+size 11108408

examples/{auto_computer_science_jobs_in_usa_20250711_205931 → auto_computer_science_jobs_in_usa_20250724_104828}/heatmaps.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a984809e4391ed6013eb04efb89d5dd0a60132fc874b78d429dde0ccf0d6b38e
-size 4647360
+oid sha256:887d8b4d5e0356a883d1c9111394f7f77c5218f09b40902a3b7cb389e5150bdf
+size 4641044

examples/{auto_graph_of_profession_choices_in_20250711_205556 → auto_graph_of_profession_choices_in_20250724_104837}/heatmaps.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e8e9974a3cbced4acef7ddfd113b4e566acc0e04a9d5206f72b8c3c869e9ee21
-size 4335264
+oid sha256:9dfa09934a3b8937c933ab8a3df57ad1b177fdb747ab004c4401bfd736e60fd9
+size 4308932

similarity.py

@@ -223,8 +223,8 @@ class JinaV4SimilarityMapper:
         image_embeddings = image_embeddings[0]
         
         # Remove special tokens
-        vision_start_position_from_start = 5
-        vision_end_position_from_end = 6
+        vision_start_position_from_start = 4
+        vision_end_position_from_end = 7
         image_embeddings = image_embeddings[vision_start_position_from_start:-vision_end_position_from_end]
         
         return pil_image, image_embeddings, size, (grid_height, grid_width)

test.py

@@ -16,13 +16,13 @@ api_key = os.getenv("JINA_TOKEN")
 MAPPER.model.set_api_key(api_key)
 
 inputs = [
-    ("A person riding a bicycle near the ocean", "https://cdn.duvine.com/wp-content/uploads/2016/04/17095703/Slides_mallorca_FOR-WEB.jpg"),
-    ("Computing jobs in the US", "https://www.ayresassociates.com/wp-content/uploads/2019/02/Career-Expo-Pie-Chart-Crop.jpg"),
-    ("Graph of profession choices in bangladesh", "https://notepadacademy.com/wp-content/uploads/2023/08/image.png")
+    ("A group of cyclists riding nearby the ocean", "https://cdn.duvine.com/wp-content/uploads/2016/04/17095703/Slides_mallorca_FOR-WEB.jpg"),
+    ("Computer Science jobs in USA", "https://www.ayresassociates.com/wp-content/uploads/2019/02/Career-Expo-Pie-Chart-Crop.jpg"),
+    ("Graph of profession choices in Bangladesh", "https://notepadacademy.com/wp-content/uploads/2023/08/image.png")
 ]
 
 for input in inputs:
-    img_proc, _, _  = MAPPER.process_image(input[1])
+    img_proc, *_ = MAPPER.process_image(input[1])
     toks, maps = MAPPER.get_token_similarity_maps(input[0], img_proc)
     save_run(input[0], input[1], img_proc, maps)