trying fa2 again

app.py

@@ -14,7 +14,7 @@ model_name = "augmxnt/shisa-7b-v1"
 
 # UI Settings
 title = "Shisa 7B"
-description = "Test out <a href='https://huggingface.co/augmxnt/shisa-7b-v1'>Shisa 7B</a> in either English or Japanese. If you aren't getting the right language outputs, you can try changing the system prompt to the appropriate language. Note, we are running `load_in_4bit` to fit in 16GB of VRAM."
+description = "Test out <a href='https://huggingface.co/augmxnt/shisa-7b-v1'>Shisa 7B</a> in either English or Japanese. If you aren't getting the right language outputs, you can try changing the system prompt to the appropriate language.\n\nNote: we are running this model quantized at `load_in_4bit` to fit in 16GB of VRAM."
 placeholder = "Type Here / ここに入力してください" 
 examples = [
     ["What are the best slices of pizza in New York City?"],
@@ -35,6 +35,7 @@ model = AutoModelForCausalLM.from_pretrained(
     device_map="auto",
     # load_in_8bit=True,
     load_in_4bit=True,
+    use_flash_attention_2=True,
 )
 
 def chat(message, history, system_prompt):