Update app.py

app.py

@@ -5,7 +5,7 @@ from dotenv import load_dotenv
 from httpx import Client
 from huggingface_hub import HfApi
 import pandas as pd
-from transformers import pipeline
+#from transformers import pipeline
 import spaces
 from llama_cpp import Llama
 
@@ -22,11 +22,13 @@ headers = {
 }
 client = Client(headers=headers)
 api = HfApi(token=HF_TOKEN)
-pipe = pipeline("text-generation", model="motherduckdb/DuckDB-NSQL-7B-v0.1", device="cuda")
+#pipe = pipeline("text-generation", model="motherduckdb/DuckDB-NSQL-7B-v0.1", device="cuda")
 llama = Llama(
         model_path="DuckDB-NSQL-7B-v0.1-q8_0.gguf",
         n_ctx=2048,
+        n_gpu_layers=50
     )
+
 @spaces.GPU
 def generate_sql(prompt):
     # pred = pipe(prompt, max_length=1000)