Update app.py

app.py

@@ -1,7 +1,6 @@
 import os
 from threading import Thread
 from typing import Iterator
-
 import gradio as gr
 import spaces
 import torch
@@ -11,15 +10,9 @@ 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
-
-This Space demonstrates text generation using different models. Choose a model from the dropdown and experience its creative capabilities!
+# 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 
 """
 
 LICENSE = """
@@ -28,25 +21,29 @@ LICENSE = """
 """
 
 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>"
+
+models_cache = {}
 
 @spaces.GPU(duration=120)
 def generate(
+    model_id: str,
     message: str,
     chat_history: list[tuple[str, str]],
     system_prompt: str,
-    model_id: str = None,  # Add default value for model_id
     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]:
-    if not model_id:
-        raise ValueError("Please select a model from the dropdown.")
-    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
+    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]
 
     conversation = []
     if system_prompt:
@@ -82,19 +79,17 @@ def generate(
         yield "".join(outputs)
 
 
-model_dropdown = gr.Dropdown(label="Select Model", choices=list(MODELS.values()))
-
 chat_interface = gr.ChatInterface(
     fn=generate,
     additional_inputs=[
-        model_dropdown,
+        gr.Textbox(label="Model ID", placeholder="Nekochu/Luminia-13B-v3"),
         gr.Textbox(label="System prompt", lines=6),
         gr.Slider(
             label="Max new tokens",
             minimum=1,
             maximum=MAX_MAX_NEW_TOKENS,
             step=1,
-            value=DEFAULT_MAX
+            value=DEFAULT_MAX_NEW_TOKENS,
         ),
         gr.Slider(
             label="Temperature",