test

app.py

@@ -19,13 +19,13 @@ model.to(device)
 
 
 def generate_image(text):
-    return "./assets/ICA-Logo.png"
+    return Image.open("./assets/ICA-Logo.png").convert("RGB")
 
 # Interface do Gradio
 iface = gr.Interface(
     fn=generate_image,
     inputs="text",
-    outputs="file",
+    outputs="image",
     live=True,
     capture_session=True,
     title="Texto para Imagem",

utils.py

@@ -0,0 +1,9 @@
+import os
+
+def calculate_similarity():
+    image_features = embeddings[ModalityType.VISION]
+    text_features = embeddings[ModalityType.TEXT]
+
+    image_features /= image_features.norm(dim=-1, keepdim=True)
+    text_features /= text_features.norm(dim=-1, keepdim=True)
+    similarity = text_features.cpu().numpy() @ image_features.cpu().numpy().T