Update app.py

app.py

@@ -1,5 +1,7 @@
 import gradio as gr
 from transformers import CLIPModel, AutoTokenizer, RawImage
+import torch
+import torch.nn.functional as F
 
 # Load the CLIP model and tokenizer
 model = CLIPModel.from_pretrained("Xenova/mobileclip_blt")
@@ -18,11 +20,9 @@ def compute_probability(image):
     text_embeds = model(text_inputs)
     normalized_text_embeds = text_embeds.normalize().tolist()
 
-    probabilities = normalized_image_embeds.map(
-        x => softmax(normalized_text_embeds.map(y => 100 * dot(x, y)))
-    )
+    probabilities = [F.softmax(torch.tensor([100 * torch.dot(torch.tensor(x), torch.tensor(y)) for y in normalized_text_embeds])).tolist()[0] for x in normalized_image_embeds]
 
-    return {"probability": probabilities[0][0]}
+    return {"probability": probabilities[0]}
 
 # Create the Gradio interface
 iface = gr.Interface(