Update app.py

app.py

@@ -31,10 +31,14 @@ from huggingface_hub import login, HfFolder
 # model = PeftModel.from_pretrained(model, lora_weights)
 # Load model directly
 # from transformers import AutoTokenizer, AutoModelForCausalLM
-
+bnb_config = BitsAndBytesConfig(
+    load_in_8bit=True,
+    llm_int8_threshold=6.0,
+)
 tokenizer = AutoTokenizer.from_pretrained("FlawedLLM/Bhashini_gemma_merged4bit_clean_final")
 model = AutoModelForCausalLM.from_pretrained("FlawedLLM/Bhashini_gemma_merged4bit_clean_final",
-                                             device_map="auto")
+                                             device_map="auto",
+                                             quantization_config=bnb_config,)
 # alpaca_prompt = You MUST copy from above!
 @spaces.GPU(duration=300)
 def chunk_it(input_command, item_list):