first commit

app.py

@@ -0,0 +1,61 @@
+import spaces
+import gradio as gr
+from transformers import pipeline
+import torch
+
+# Load the Gemma 3 pipeline.
+# Gemma 3 is a multimodal model that accepts text and image inputs.
+pipe = pipeline(
+    "image-text-to-text",
+    model="google/gemma-3-4b-it",
+    device="cuda" if torch.cuda.is_available() else "cpu",
+    torch_dtype=torch.bfloat16 if torch.cuda.is_available() else None
+)
+@spaces.GPU
+def generate_response(user_text, user_image, history):
+    """
+    This function takes the user's text and an optional image,
+    creates a message list (including a system message), and then
+    uses the Gemma 3 pipeline to generate a response.
+    """
+    # Start with a system prompt.
+    messages = [
+        {
+            "role": "system",
+            "content": [{"type": "text", "text": "You are a helpful assistant."}]
+        }
+    ]
+    # Build user content from the provided image and/or text.
+    user_content = []
+    if user_image is not None:
+        # The image input is a PIL image.
+        user_content.append({"type": "image", "image": user_image})
+    if user_text:
+        user_content.append({"type": "text", "text": user_text})
+    messages.append({"role": "user", "content": user_content})
+    
+    # Generate output using the model pipeline.
+    output = pipe(text=messages, max_new_tokens=200)
+    # Extract the generated text.
+    # (This extraction follows the code snippet provided in the model card.)
+    response = output[0][0]["generated_text"][-1]["content"]
+    
+    # Append the new exchange to the conversation history.
+    history.append((user_text, response))
+    return history, history
+
+with gr.Blocks() as demo:
+    gr.Markdown("# Gemma 3 Chat Interface")
+    gr.Markdown(
+        "This interface lets you chat with the Gemma 3 model. "
+        "You can type a message and optionally attach an image."
+    )
+    chatbot = gr.Chatbot()
+    with gr.Row():
+        txt = gr.Textbox(show_label=False, placeholder="Type your message here...", container=False)
+        img = gr.Image(source="upload", tool="editor", type="pil", label="Attach an image (optional)")
+    state = gr.State([])
+
+    # When the user submits text (and possibly an image), call generate_response.
+    txt.submit(generate_response, inputs=[txt, img, state], outputs=[chatbot, state])
+    demo.launch()