attempt10 fix

app.py

@@ -11,37 +11,50 @@ MAX_MAX_NEW_TOKENS = 2048
 DEFAULT_MAX_NEW_TOKENS = 1024
 MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
 
-MODELS = {
-    "Nekochu/Luminia-13B-v3": "Default - Nekochu/Luminia-13B-v3",
-    "Nekochu/Llama-2-13B-German-ORPO": "German ORPO - Nekochu/Llama-2-13B-German-ORPO",
-}
-
 DESCRIPTION = """\
-# Text Generation with Selectable Models
+# Nekochu/Luminia-13B-v3
+This Space demonstrates model Nekochu/Luminia-13B-v3 by Nekochu, a Llama 2 model with 13B parameters fine-tuned for SD gen prompt 
+"""
 
-This Space demonstrates text generation using different models. Choose a model from the dropdown and experience its creative capabilities!
+LICENSE = """
+<p/>
+---.
 """
 
-LICENSE = """<p/> ---."""
+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>"
+    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, tokenizer = load_model(model_id)
+
+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"},
+]
 
 @spaces.GPU(duration=120)
 def generate(
+    model_dropdown: str,
+    custom_model_id: str,
     message: str,
     chat_history: list[tuple[str, str]],
     system_prompt: str,
-    model_id: str,
     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 = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", load_in_4bit=True)
-    tokenizer = AutoTokenizer.from_pretrained(model_id)
-    tokenizer.use_default_system_prompt = False
+    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})
@@ -75,11 +88,18 @@ def generate(
         outputs.append(text)
         yield "".join(outputs)
 
-model_dropdown = gr.Dropdown(label="Select Model", choices=list(MODELS.values()))
+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=[
         model_dropdown,
+        custom_model_id_input,
         gr.Textbox(label="System prompt", lines=6),
         gr.Slider(
             label="Max new tokens",
@@ -131,4 +151,4 @@ 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()