code added

app.py

@@ -0,0 +1,103 @@
+"""
+Gradio demo for UI‑TARS 1.5‑7B (image‑text‑to‑text) on Hugging Face Spaces.
+Save this file as **app.py** and add a *requirements.txt* with the packages
+listed below.  Then create a new **Python** Space, upload both files and
+commit — the Space will build and serve the app automatically.
+
+requirements.txt (suggested versions)
+-------------------------------------
+transformers==4.41.0
+accelerate>=0.29.0
+torch>=2.2
+sentencepiece  # needed for many multilingual models
+bitsandbytes    # optional: enables 4‑bit quantization if Space has GPU
+pillow
+gradio>=4.33
+"""
+
+from __future__ import annotations
+
+from typing import List, Dict, Any
+
+import gradio as gr
+from PIL import Image
+from transformers import pipeline
+import base64
+
+def load_model():
+    """Load the UI‑TARS multimodal pipeline once at startup."""
+    print("Loading UI‑TARS 1.5‑7B… this may take a while the first time.")
+    return pipeline(
+        "image-text-to-text",
+        model="ByteDance-Seed/UI-TARS-1.5-7B",
+        device_map="auto",  # automatically use GPU if available
+    )
+
+
+pipe = load_model()
+
+
+def answer_question(image: Image.Image, question: str) -> str:
+    """Run the model on the provided image & question and return its answer."""
+    if image is None or not question.strip():
+        return "Please supply **both** an image and a question."
+    
+    base64_image = base64.b64encode(image.tobytes()).decode('utf-8')
+
+    # Compose a messages list in the expected multimodal chat format.
+    messages: List[Dict[str, Any]] = [
+        {
+            "role": "user",
+            "content": [
+                {"type": "text", "text": f"You are a GUI agent. You are given a task and your action history, with screenshots. You need to perform the next action to complete the task. \n\n## Output Format\n```\nThought: ...\nAction: ...\n```\n\n## Action Space\n\nclick(start_box='<|box_start|>(x1, y1)<|box_end|>')\nleft_double(start_box='<|box_start|>(x1, y1)<|box_end|>')\nright_single(start_box='<|box_start|>(x1, y1)<|box_end|>')\ndrag(start_box='<|box_start|>(x1, y1)<|box_end|>', end_box='<|box_start|>(x3, y3)<|box_end|>')\nhotkey(key='')\ntype(content='') #If you want to submit your input, use \"\\n\" at the end of `content`.\nscroll(start_box='<|box_start|>(x1, y1)<|box_end|>', direction='down or up or right or left')\nwait() #Sleep for 5s and take a screenshot to check for any changes.\nfinished(content='xxx') # Use escape characters \\', \\\", and \\n in content part to ensure we can parse the content in normal python string format.\n\n\n## Note\n- Use Chinese in `Thought` part.\n- Write a small plan and finally summarize your next action (with its target element) in one sentence in `Thought` part.\n\n## User Instruction\n{question.strip()}"},
+            ],
+        },
+        {
+            "role":"user",
+            "content": [
+                {"type": "image_url", 
+                "image_url": base64_image},
+            ],
+        }
+    ]
+
+    # The pipeline returns a list with one dict when `messages` is passed via
+    # the `text` keyword. We extract the generated text robustly.
+    outputs = pipe(text=messages)
+
+    if isinstance(outputs, list):
+        first = outputs[0]
+        if isinstance(first, dict) and "generated_text" in first:
+            return first["generated_text"].strip()
+        return str(first)
+
+    return str(outputs)
+
+
+demo = gr.Interface(
+    fn=answer_question,
+    inputs=[
+        gr.Image(type="pil", label="Upload image"),
+        gr.Textbox(label="Ask a question about the image", placeholder="e.g. What animal is on the candy?"),
+    ],
+    outputs=gr.Textbox(label="UI‑TARS answer"),
+    title="UI‑TARS 1.5‑7B – Visual Q&A",
+    description=(
+        "Upload an image and ask a question. The **UI‑TARS 1.5‑7B** model will "
+        "answer based on the visual content. Runs completely on‑device in this Space."
+    ),
+    examples=[
+        [
+            "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG",
+            "What animal is on the candy?",
+        ]
+    ],
+    cache_examples=True,
+    allow_flagging="never",
+)
+
+
+if __name__ == "__main__":
+    # Spaces automatically call `demo.launch()`, but running locally this
+    # guard lets you execute `python app.py` for quick tests.
+    demo.launch()

requirements.txt

@@ -0,0 +1,7 @@
+transformers
+accelerate
+torch
+sentencepiece
+bitsandbytes
+pillow
+gradio