Try to load the model in 8 bits

app.py

@@ -7,7 +7,7 @@ import re
 logging.basicConfig(level=logging.INFO)
 logging.getLogger('transformers').setLevel(logging.INFO)
 
-llama = pipeline("text-generation", model="filipealmeida/open-llama-3b-v2-pii-transform")
+llama = pipeline("text-generation", model="filipealmeida/open-llama-3b-v2-pii-transform", model_kwargs={"load_in_8bit": True})
 
 def generate_text(prompt, example):  
     logging.debug(f"Received prompt: {prompt}")

requirements.txt

@@ -1,2 +1,4 @@
 transformers==4.31.0
-torch==2.0.1
+torch==2.0.1
+bitsandbytes
+accelerate