Update app.py

app.py

@@ -1,17 +1,23 @@
 import gradio as gr
 import spaces
-from transformers import pipeline
 import torch
+from transformers import BitsAndBytesConfig, pipeline
+
+quant_config = BitsAndBytesConfig(
+    load_in_4bit=True,
+    bnb_4bit_quant_type="nf4",
+    bnb_4bit_compute_dtype=torch.float16,
+    bnb_4bit_use_double_quant=True
+)
 
-# Initialize model with explicit dtype
 model = pipeline(
     "text-generation",
     model="unsloth/DeepSeek-R1-Distill-Llama-8B",
-    device_map="auto",
-    torch_dtype=torch.float16,  # Now recognizes 'torch'
-    model_kwargs={"load_in_8bit": True}
+    quantization_config=quant_config,
+    device_map="auto"
 )
 
+
 @spaces.GPU(duration=300)  # Increased to 5 minutes
 def chat_response(message, history):
     # Add explicit initialization check