Update app.py

app.py

@@ -49,15 +49,12 @@ def load_models():
     model_name = "deepseek-ai/DeepSeek-R1-Distill-Llama-8B"
     app.state.deepseek_tokenizer = AutoTokenizer.from_pretrained(model_name)
     app.state.deepseek_tokenizer.pad_token = app.state.deepseek_tokenizer.eos_token
-    
+
     app.state.deepseek_model = AutoModelForCausalLM.from_pretrained(
         model_name,
-        torch_dtype="auto",
-        device_map="auto",
-        offload_folder="offload"   # folder to store offloaded weights
+        load_in_8bit=True, # requires bitsandbytes device_map="auto"
     )
 
-
 # -----------------------------
 # Model handlers
 # -----------------------------