Adding a simple monkey search for Leetcode - Darn LeetMonkey

app.py

@@ -23,7 +23,7 @@ dense_model = SentenceTransformer('sentence-transformers/all-Mpnet-base-v2', dev
 # Load the quantized Llama 2 model and tokenizer
 model_name = "TheBloke/Llama-2-7B-Chat-GPTQ"
 tokenizer = AutoTokenizer.from_pretrained(model_name)
-model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto")
+model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto", quantization_config={"disable_exllama": True})
 
 def search_problems(query, top_k=5):
     dense_query = dense_model.encode([query])[0].tolist()