asda

app.py

@@ -19,8 +19,6 @@ classifier = joblib.load(model_dir + "/finetuned_classifier.pkl")
 embedding_model = mr.get_model("news_embedding", version = 1)
 model_dir = embedding_model.download()
 embedding_model = joblib.load(model_dir + "/news_embedding.pkl")
-embedding_model = embedding_model.to(device)
-
 
 index_to_category = {
     0:"Polititcs",
@@ -59,7 +57,7 @@ The input is currently truncated at around 400 words so make sure to include the
 
 
 def predict(text):
-    embedding = embedding_model.encode([text])
+    embedding = embedding_model.encode([text], device = device)
     with torch.no_grad():
         embedding = torch.tensor(embedding, device=device, dtype=torch.float32)
         probs = classifier.probabilities(embedding).cpu().numpy()