np array to dict

interfaces/cap_minor_media.py

@@ -95,7 +95,11 @@ def predict(text, major_model_id, minor_model_id, tokenizer_id, HF_TOKEN=None):
     top_major_id = major_index_to_id[top_major_index]
 
     # Default: show major topic predictions
-    filtered_probs = normalize_probs(np.argsort(major_probs_np)[::-1])
+    filtered_probs = {
+        i: float(major_probs_np[i])
+        for i in np.argsort(major_probs_np)[::-1]
+    }
+    filtered_probs = normalize_probs(filtered_probs)
     output_pred = {
         f"[{major_index_to_id[k]}] {CAP_MEDIA_LABEL_NAMES[filtered_probs[k]]}": v
         for k, v in sorted(filtered_probs.items(), key=lambda item: item[1], reverse=True)
@@ -135,7 +139,7 @@ def predict_cap(text, language, domain):
     return predict(text, major_model_id, minor_model_id, tokenizer_id)
 
 demo = gr.Interface(
-    title="CAP Media Topics Babel Demo",
+    title="CAP Media/Minor Topics Babel Demo",
     fn=predict_cap,
     inputs=[gr.Textbox(lines=6, label="Input"),
             gr.Dropdown(languages, label="Language"),