Update llm.py

llm.py

@@ -196,7 +196,7 @@ def ask_ai(
         f16_kv=True,
         max_tokens=200
     )
-    embeddings = HuggingFaceEmbeddings()
+    embeddings = HuggingFaceEmbeddings(model_kwargs={"device": "cuda"})
     embed_model = LangchainEmbedding(embeddings)
     llm_predictor = LLMPredictor(
         llm=llm