Update app.py

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 = """
@@ -21,22 +21,29 @@ 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)
 
-models_cache = {}
+MODELS = [
+    {"name": "Nekochu/Luminia-13B-v3", "id": "Nekochu/Luminia-13B-v3"},
+    {"name": "Nekochu/Llama-2-13B-German-ORPO", "id": "Nekochu/Llama-2-13B-German-ORPO"},
+    # Add more models here in the future
+]
 
 @spaces.GPU(duration=120)
 def generate(
-    model_id: str,
+    model_dropdown: str,
+    custom_model_id: str,
     message: str,
     chat_history: list[tuple[str, str]],
     system_prompt: str,
@@ -46,13 +53,11 @@ def generate(
     top_k: int = 50,
     repetition_penalty: float = 1.2,
 ) -> Iterator[str]:
-    if model_id not in models_cache:
-        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
-        models_cache[model_id] = (model, tokenizer)
-    else:
-        model, tokenizer = models_cache[model_id]
+    # Prioritize custom model ID if provided, otherwise use the dropdown selection
+    selected_model_id = custom_model_id if custom_model_id else model_dropdown
+    model, tokenizer = load_model(selected_model_id)
+
+    conversation = []
     if system_prompt:
         conversation.append({"role": "system", "content": system_prompt})
     for user, assistant in chat_history:
@@ -85,11 +90,18 @@ def generate(
         outputs.append(text)
         yield "".join(outputs)
 
+model_dropdown = gr.Dropdown(
+    label="Select Predefined Model",
+    choices=[model["name"] for model in MODELS],
+    value=MODELS[0]["name"], # Default to the first model
+)
+custom_model_id_input = gr.Textbox(label="Or Enter Custom Model ID", placeholder="Enter model ID here")
 
 chat_interface = gr.ChatInterface(
     fn=generate,
     additional_inputs=[
-        gr.Textbox(label="Model ID", placeholder="Nekochu/Luminia-13B-v3"),
+        model_dropdown,
+        custom_model_id_input,
         gr.Textbox(label="System prompt", lines=6),
         gr.Slider(
             label="Max new tokens",
@@ -141,4 +153,4 @@ 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()