Revert back fix2

app.py

@@ -13,7 +13,7 @@ MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
 
 DESCRIPTION = """\
 # Nekochu/Luminia-13B-v3
-This Space demonstrates model [Nekochu/Luminia-13B-v3](https://huggingface.co/Nekochu/Luminia-13B-v3) by Nekochu, a Llama 2 model with 13B parameters fine-tuned for SD gen prompt 
+This Space demonstrates model Nekochu/Luminia-13B-v3 by Nekochu, a Llama 2 model with 13B parameters fine-tuned for SD gen prompt 
 """
 
 LICENSE = """
@@ -21,15 +21,18 @@ LICENSE = """
 ---.
 """
 
+def load_model(model_id):
+    model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", load_in_4bit=True)
+    tokenizer = AutoTokenizer.from_pretrained(model_id)
+    tokenizer.use_default_system_prompt = False
+    return model, tokenizer
+
 if not torch.cuda.is_available():
     DESCRIPTION += "\n<p>Running on CPU 🥶 This demo does not work on CPU.</p>"
 
-
 if torch.cuda.is_available():
     model_id = "Nekochu/Luminia-13B-v3"
-    model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", load_in_4bit=True)
-    tokenizer = AutoTokenizer.from_pretrained(model_id)
-    tokenizer.use_default_system_prompt = False
+    model, tokenizer = load_model(model_id)
 
 
 @spaces.GPU(duration=120)
@@ -37,12 +40,14 @@ def generate(
     message: str,
     chat_history: list[tuple[str, str]],
     system_prompt: str,
+    model_id: str = "Nekochu/Luminia-13B-v3",
     max_new_tokens: int = 1024,
     temperature: float = 0.6,
     top_p: float = 0.9,
     top_k: int = 50,
     repetition_penalty: float = 1.2,
 ) -> Iterator[str]:
+    model, tokenizer = load_model(model_id)
     conversation = []
     if system_prompt:
         conversation.append({"role": "system", "content": system_prompt})
@@ -81,6 +86,7 @@ chat_interface = gr.ChatInterface(
     fn=generate,
     additional_inputs=[
         gr.Textbox(label="System prompt", lines=6),
+        gr.Textbox(label="Model ID", placeholder="Nekochu/Luminia-13B-v3"),
         gr.Slider(
             label="Max new tokens",
             minimum=1,
@@ -131,4 +137,4 @@ with gr.Blocks(css="style.css") as demo:
     gr.Markdown(LICENSE)
 
 if __name__ == "__main__":
-    demo.queue(max_size=20).launch()
+    demo.queue(max_size=20).launch()