Single language

app.py

@@ -1049,9 +1049,9 @@ if not precomputed_results:
     model = model.to(device)
 
 
-def change_language(lang, randomize_imgs, randomize_labels):
+def change_language(randomize_imgs, randomize_labels):
     # compute text embeddings
-    labels = babel_imagenet[lang][1]
+    labels = babel_imagenet["EN"][1]
     class_order = list(range(len(labels)))
     if randomize_labels:
         np.random.shuffle(class_order)
@@ -1065,7 +1065,7 @@ def change_language(lang, randomize_imgs, randomize_labels):
     else:
         text_features = None
     correct_text = gr.Text(
-        f"Correct was: ''. Question 1/{len(babel_imagenet[lang][0])} ", label="Game"
+        f"Correct was: ''. Question 1/{len(babel_imagenet["EN"][0])} ", label="Game"
     )
     player_score_text = gr.Text(f"Your choice: (Score: 0) ", label="Player")
     clip_score_text = gr.Text(f"mSigLIP chose: '' (Score: 0)", label="Opponent")
@@ -1082,7 +1082,7 @@ def change_language(lang, randomize_imgs, randomize_labels):
     )
 
 
-def select(idx, lang, choice, correct, model_choice, player_score, clip_score, choices):
+def select(idx, choice, correct, model_choice, player_score, clip_score, choices):
     # checks if answer choice is correct and updated scores
     correct_name, correct_value = correct
     model_choice_name, model_choice_value = model_choice
@@ -1095,7 +1095,7 @@ def select(idx, lang, choice, correct, model_choice, player_score, clip_score, c
     clip_score = clip_score + int(model_correct)
 
     correct_text = gr.Text(
-        f"Correct was: '{correct_name}'. Question {idx+1}/{len(babel_imagenet[lang][0])} ",
+        f"Correct was: '{correct_name}'. Question {idx+1}/{len(babel_imagenet["EN"][0])} ",
         label="Game",
     )
     player_score_text = gr.Text(
@@ -1110,19 +1110,19 @@ def select(idx, lang, choice, correct, model_choice, player_score, clip_score, c
     return correct_text, player_score_text, clip_score_text, player_score, clip_score
 
 
-def prepare(raw_idx, lang, text_embeddings, class_order, randomize_images):
+def prepare(raw_idx, text_embeddings, class_order, randomize_images):
     #  prepared next question, loads image, and computes choices
 
-    raw_idx = (raw_idx + 1) % len(babel_imagenet[lang][0])
+    raw_idx = (raw_idx + 1) % len(babel_imagenet["EN"][0])
     idx = class_order[raw_idx]
-    lang_class_idxs = babel_imagenet[lang][0]
+    lang_class_idxs = babel_imagenet["EN"][0]
     class_idx = lang_class_idxs[idx]
 
     # skip classes with no images
     while class_idx in no_image_idxs:
-        raw_idx = (raw_idx + 1) % len(babel_imagenet[lang][0])
+        raw_idx = (raw_idx + 1) % len(babel_imagenet["EN"][0])
         idx = class_order[raw_idx]
-        lang_class_idxs = babel_imagenet[lang][0] if lang != "EN" else list(range(1000))
+        lang_class_idxs = babel_imagenet["EN"][0] if "EN" != "EN" else list(range(1000))
         class_idx = lang_class_idxs[idx]
 
     img_idx = 0
@@ -1131,7 +1131,7 @@ def prepare(raw_idx, lang, text_embeddings, class_order, randomize_images):
             min(len(babelnet_images[class_idx]), max_image_choices)
         )
     img_url = babelnet_images[class_idx][img_idx]["url"]
-    class_labels = babel_imagenet[lang][1] if lang != "EN" else openai_en_classes
+    class_labels = babel_imagenet["EN"][1] if "EN" != "EN" else openai_en_classes
 
     if not precomputed_results:
         try:
@@ -1150,14 +1150,14 @@ def prepare(raw_idx, lang, text_embeddings, class_order, randomize_images):
         except:
             gr.Warning("There is a problem with the next class. Skipping it.")
             return prepare(
-                raw_idx, lang, text_embeddings, class_order, randomize_images
+                raw_idx, text_embeddings, class_order, randomize_images
             )
 
         similarity = (text_embeddings @ image_features.cpu().numpy().T).squeeze()
         choices = np.argsort(similarity)[-4:].tolist()
     else:
         choices = list(
-            reversed(precomputed_results[lang][idx][img_idx])
+            reversed(precomputed_results["EN"][idx][img_idx])
         )  # precomputing script uses torch.topk which sorts in reverse here
     if idx not in choices:
         choices = [idx] + choices[1:]
@@ -1194,11 +1194,11 @@ def prepare(raw_idx, lang, text_embeddings, class_order, randomize_images):
     return next_radio, next_image, raw_idx, correct_choice, model_choice, choice_values
 
 
-def reroll(raw_idx, lang, text_embeddings, class_order, randomize_images):
+def reroll(raw_idx, text_embeddings, class_order, randomize_images):
     #  prepared next question, loads image, and computes choices
 
     idx = class_order[raw_idx]
-    lang_class_idxs = babel_imagenet[lang][0]
+    lang_class_idxs = babel_imagenet["EN"][0]
     class_idx = lang_class_idxs[idx]
 
     img_idx = 0
@@ -1207,7 +1207,7 @@ def reroll(raw_idx, lang, text_embeddings, class_order, randomize_images):
             min(len(babelnet_images[class_idx]), max_image_choices)
         )
     img_url = babelnet_images[class_idx][img_idx]["url"]
-    class_labels = babel_imagenet[lang][1] if lang != "EN" else openai_en_classes
+    class_labels = babel_imagenet["EN"][1] if "EN" != "EN" else openai_en_classes
 
     if not precomputed_results:
         try:
@@ -1226,14 +1226,14 @@ def reroll(raw_idx, lang, text_embeddings, class_order, randomize_images):
         except:
             gr.Warning("There is a problem with the next class. Skipping it.")
             return prepare(
-                raw_idx, lang, text_embeddings, class_order, randomize_images
+                raw_idx, text_embeddings, class_order, randomize_images
             )
 
         similarity = (text_embeddings @ image_features.cpu().numpy().T).squeeze()
         choices = np.argsort(similarity)[-4:].tolist()
     else:
         choices = list(
-            reversed(precomputed_results[lang][idx][img_idx])
+            reversed(precomputed_results["EN"][idx][img_idx])
         )  # precomputing script uses torch.topk which sorts in reverse here
     if idx not in choices:
         choices = [idx] + choices[1:]
@@ -1390,7 +1390,6 @@ Select your language, click 'Start' and start guessing! We'll keep track of your
         fn=select,
         inputs=[
             class_idx,
-            "EN",
             options,
             correct_choice,
             model_choice,
@@ -1409,7 +1408,6 @@ Select your language, click 'Start' and start guessing! We'll keep track of your
         fn=prepare,
         inputs=[
             class_idx,
-            "EN",
             text_embeddings,
             class_order,
             randomize_images,
@@ -1434,7 +1432,6 @@ Select your language, click 'Start' and start guessing! We'll keep track of your
         fn=prepare,
         inputs=[
             class_idx,
-            "EN",
             text_embeddings,
             class_order,
             randomize_images,
@@ -1446,7 +1443,6 @@ Select your language, click 'Start' and start guessing! We'll keep track of your
         fn=reroll,
         inputs=[
             class_idx,
-            "EN",
             text_embeddings,
             class_order,
             randomize_images,