use spaces gpu

app.py

@@ -114,6 +114,7 @@ def load_model():
     generator_mini = pipeline(task="text-generation", model=ankerbot_model, tokenizer=ankerbot_tokenizer, torch_dtype=torch.float16, trust_remote_code=True) # True for flash-attn2 else False
 
 load_model()
+@spaces.GPU
 def generate_response(query, context, prompts, max_tokens, temperature, top_p):
     system_message_support = f"""<|im_start|>system
         Rolle: Du bist der KI-Assistent für Kundenservice, der im Namen des Unternehmens und Gewürzmanufaktur Ankerkraut handelt und Antworten aus der Ich-Perspektive, basierend auf den bereitgestellten Informationen gibt.
@@ -175,12 +176,12 @@ def generate_response(query, context, prompts, max_tokens, temperature, top_p):
         response = response.split("assistant").pop().strip()
 
     return response
-
+@spaces.GPU
 def get_embedding(text):
     """Generate an embedding using Sentence Transformers."""
     embedding = model.encode(text, normalize_embeddings=True)  # Normalize for cosine similarity
     return embedding
-
+@spaces.GPU
 def search_qdrant_with_context(query_text, collection_name, top_k=3):
     """Search Qdrant using a GPT-2 generated embedding."""
     query_embedding = get_embedding(query_text)  # Convert prompt to embedding
@@ -197,6 +198,7 @@ def search_qdrant_with_context(query_text, collection_name, top_k=3):
     print("Retrieved Text ", retrieved_texts)
 
     return retrieved_texts
+@spaces.GPU
 def respond(
     query,
     history: list[tuple[str, str]],