create the app.py file

app.py

@@ -0,0 +1,48 @@
+import gradio as gr
+from transformers import pipeline, AutoTokenizer
+
+model_names = [
+    "distilgpt2",
+    "gpt2",
+]
+
+# --- minimal caching so the model isn't reloaded every click ---
+_pipe_cache = {}
+
+def _get_pipe(model_name):
+    if model_name not in _pipe_cache:
+        tok = AutoTokenizer.from_pretrained(model_name)
+        _pipe_cache[model_name] = pipeline(
+            "text-generation",
+            model=model_name,
+            tokenizer=tok,
+            device_map="auto",
+            torch_dtype="auto",
+        )
+    return _pipe_cache[model_name]
+# ---------------------------------------------------------------
+
+def generate_with_choice(prompt, model_name):
+    pipe = _get_pipe(model_name)
+    out = pipe(
+        prompt,
+        max_new_tokens=50,
+        do_sample=True,
+        return_full_text=False
+    )
+    return out[0]["generated_text"]
+
+demo2 = gr.Interface(
+    fn=generate_with_choice,
+    inputs=[
+        gr.Textbox(lines=4, label="Enter Prompt"),
+        gr.Dropdown(model_names, label="Choose Model"),
+    ],
+    outputs=gr.Textbox(lines=5, label="Output"),
+    flagging_mode="never",
+    title="Model Chooser Demo",
+    description="Pick a model and generate text on the fly!",
+    theme="soft",
+)
+
+demo2.queue().launch(share=True)