Update app.py

app.py

@@ -1,12 +1,94 @@
 import gradio as gr
+import torch
+from threading import Thread
+from transformers import (
+    AutoProcessor,
+    AutoModelForImageTextToText,
+    TextIteratorStreamer,
+)
 
-def hello(x):
-    return f"Hello {x}"
+# ======================
+# INIT
+# ======================
+DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+MODEL_ID = "HuggingFaceTB/SmolVLM2-2.2B-Instruct"
 
-demo = gr.Interface(
-    fn=hello,
-    inputs=gr.Textbox(),
-    outputs=gr.Textbox(),
-)
+processor = AutoProcessor.from_pretrained(MODEL_ID)
+model = AutoModelForImageTextToText.from_pretrained(
+    MODEL_ID,
+    torch_dtype=torch.bfloat16 if torch.cuda.is_available() else torch.float32,
+).to(DEVICE).eval()
+
+
+# ======================
+# STREAMING INFERENCE (SAFE)
+# ======================
+def analyze_stream(text, image, max_tokens):
+    content = []
+    if image is not None:
+        content.append({"type": "image", "path": image})
+    if text.strip():
+        content.append({"type": "text", "text": text})
+
+    messages = [{"role": "user", "content": content}]
+
+    inputs = processor.apply_chat_template(
+        messages,
+        add_generation_prompt=True,
+        tokenize=True,
+        return_tensors="pt",
+    ).to(DEVICE)
+
+    streamer = TextIteratorStreamer(
+        processor,
+        skip_prompt=True,
+        skip_special_tokens=True,
+    )
+
+    thread = Thread(
+        target=model.generate,
+        kwargs=dict(
+            **inputs,
+            streamer=streamer,
+            max_new_tokens=max_tokens,
+            do_sample=False,
+            temperature=0.0,
+        ),
+    )
+    thread.start()
+
+    partial = ""
+    for token in streamer:
+        partial += token
+        yield partial
+
+
+# ======================
+# UI STABLE
+# ======================
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.Markdown("## ⚡ SmolVLM2 – Real-time Analysis")
+
+    with gr.Row():
+        with gr.Column():
+            txt = gr.Textbox(
+                label="Question",
+                lines=3,
+            )
+            img = gr.Image(type="filepath", label="Image")
+            max_tokens = gr.Slider(50, 400, value=200, step=50)
+            btn = gr.Button("🚀 Analyze", variant="primary")
+
+        with gr.Column():
+            out = gr.Textbox(
+                label="Streaming Output",
+                lines=14,
+            )
+
+    btn.click(
+        fn=analyze_stream,
+        inputs=[txt, img, max_tokens],
+        outputs=out,
+    )
 
 demo.launch()