Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -5,71 +5,130 @@ import torch
|
|
5 |
from PIL import Image, ImageDraw
|
6 |
import gradio as gr
|
7 |
|
8 |
-
from transformers import Qwen2_5_VLForConditionalGeneration
|
9 |
-
from transformers import AutoProcessor
|
10 |
from qwen_vl_utils import process_vision_info # include this file in your repo if not pip-installable
|
11 |
|
12 |
# ---- model & processor loaded on CPU ----
|
13 |
model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
14 |
"ByteDance-Seed/UI-TARS-1.5-7B",
|
15 |
device_map="auto",
|
16 |
-
torch_dtype=torch.float32, # CPU
|
17 |
)
|
18 |
processor = AutoProcessor.from_pretrained(
|
19 |
"ByteDance-Seed/UI-TARS-1.5-7B",
|
20 |
size={"shortest_edge": 256 * 28 * 28, "longest_edge": 1344 * 28 * 28},
|
21 |
use_fast=True,
|
22 |
-
|
23 |
)
|
24 |
|
25 |
-
|
|
|
|
|
26 |
img = image.copy()
|
27 |
if point:
|
28 |
x, y = point[0] * img.width, point[1] * img.height
|
29 |
ImageDraw.Draw(img).ellipse(
|
30 |
-
(x - radius, y - radius, x + radius, y + radius), fill=
|
31 |
)
|
32 |
return img
|
33 |
|
|
|
34 |
@spaces.GPU
|
35 |
-
def navigate(
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
40 |
]
|
41 |
-
|
42 |
-
|
|
|
|
|
|
|
|
|
|
|
43 |
images, videos = process_vision_info(messages)
|
44 |
-
inputs = processor(
|
45 |
-
|
|
|
|
|
|
|
|
|
|
|
46 |
|
47 |
-
# generate
|
48 |
generated = model.generate(**inputs, max_new_tokens=128)
|
49 |
-
trimmed = [
|
50 |
-
|
|
|
|
|
|
|
|
|
51 |
|
52 |
-
#
|
53 |
try:
|
54 |
-
actions = ast.literal_eval(
|
55 |
for act in actions if isinstance(actions, list) else [actions]:
|
56 |
-
pos = act.get(
|
57 |
-
if pos and isinstance(pos, list) and len(pos)==2:
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
|
|
|
|
62 |
|
63 |
|
|
|
|
|
64 |
demo = gr.Interface(
|
65 |
fn=navigate,
|
66 |
inputs=[
|
67 |
gr.Image(type="pil", label="Screenshot"),
|
68 |
-
gr.Textbox(
|
|
|
|
|
|
|
|
|
69 |
gr.Dropdown(choices=["web", "phone"], value="web", label="Platform"),
|
|
|
|
|
|
|
|
|
|
|
70 |
],
|
71 |
-
|
72 |
-
title="ShowUI-2B Navigation Demo",
|
73 |
)
|
74 |
|
75 |
demo.launch(
|
@@ -77,4 +136,4 @@ demo.launch(
|
|
77 |
server_port=7860,
|
78 |
share=False, # or True if you need a public link
|
79 |
ssr_mode=False, # turn off experimental SSR so the process blocks
|
80 |
-
)
|
|
|
5 |
from PIL import Image, ImageDraw
|
6 |
import gradio as gr
|
7 |
|
8 |
+
from transformers import Qwen2_5_VLForConditionalGeneration, AutoProcessor
|
|
|
9 |
from qwen_vl_utils import process_vision_info # include this file in your repo if not pip-installable
|
10 |
|
11 |
# ---- model & processor loaded on CPU ----
|
12 |
model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
13 |
"ByteDance-Seed/UI-TARS-1.5-7B",
|
14 |
device_map="auto",
|
15 |
+
torch_dtype=torch.float32, # CPUβfriendly
|
16 |
)
|
17 |
processor = AutoProcessor.from_pretrained(
|
18 |
"ByteDance-Seed/UI-TARS-1.5-7B",
|
19 |
size={"shortest_edge": 256 * 28 * 28, "longest_edge": 1344 * 28 * 28},
|
20 |
use_fast=True,
|
|
|
21 |
)
|
22 |
|
23 |
+
|
24 |
+
def draw_point(image: Image.Image, point=None, radius: int = 5):
|
25 |
+
"""Overlay a red dot on the screenshot where the model clicked."""
|
26 |
img = image.copy()
|
27 |
if point:
|
28 |
x, y = point[0] * img.width, point[1] * img.height
|
29 |
ImageDraw.Draw(img).ellipse(
|
30 |
+
(x - radius, y - radius, x + radius, y + radius), fill="red"
|
31 |
)
|
32 |
return img
|
33 |
|
34 |
+
|
35 |
@spaces.GPU
|
36 |
+
def navigate(screenshot, task: str, platform: str, history):
|
37 |
+
"""Run one inference step on the GUIβreasoning model.
|
38 |
+
|
39 |
+
Args:
|
40 |
+
screenshot (PIL.Image): Latest UI screenshot.
|
41 |
+
task (str): Naturalβlanguage task description.
|
42 |
+
platform (str): Either "web" or "phone" for prompt conditioning.
|
43 |
+
history (list | str | None): Previous messages list. Accepts either an
|
44 |
+
actual Python list (via gr.JSON) or a JSON/Pythonβliteral string.
|
45 |
+
"""
|
46 |
+
|
47 |
+
# βββββββββββββββββββββ normalise history input ββββββββββββββββββββββββββ
|
48 |
+
if history in (None, ""):
|
49 |
+
history_list = []
|
50 |
+
else:
|
51 |
+
if isinstance(history, str):
|
52 |
+
try:
|
53 |
+
history_list = ast.literal_eval(history)
|
54 |
+
except Exception as exc:
|
55 |
+
raise ValueError("`history` must be a JSON/Python list: " + str(exc))
|
56 |
+
else:
|
57 |
+
history_list = history
|
58 |
+
|
59 |
+
if not isinstance(history_list, list):
|
60 |
+
raise ValueError("`history` must be a list of messages.")
|
61 |
+
|
62 |
+
# βββββββββββββββββββ construct current user message βββββββββββββββββββββ
|
63 |
+
prompt_header = (
|
64 |
+
"You are a GUI agent. You are given a task and your action history, "
|
65 |
+
"with screenshots. You need to perform the next action to complete "
|
66 |
+
"the task.\n\n## Output Format\n```\nThought: ...\nAction: ...\n```\n\n"
|
67 |
+
"## Action Space\nclick(start_box='...') / type(...)\n\n"
|
68 |
+
f"### Task\n{task}"
|
69 |
+
)
|
70 |
+
|
71 |
+
current_content = [
|
72 |
+
{"type": "text", "text": prompt_header},
|
73 |
+
{"type": "image_url", "image_url": screenshot},
|
74 |
]
|
75 |
+
|
76 |
+
messages = history_list + [{"role": "user", "content": current_content}]
|
77 |
+
|
78 |
+
# βββββββββββββββββββββββββββ model forward βββββββββββββββββββββββββββββ
|
79 |
+
text = processor.apply_chat_template(
|
80 |
+
messages, tokenize=False, add_generation_prompt=True
|
81 |
+
)
|
82 |
images, videos = process_vision_info(messages)
|
83 |
+
inputs = processor(
|
84 |
+
text=[text],
|
85 |
+
images=images,
|
86 |
+
videos=videos,
|
87 |
+
padding=True,
|
88 |
+
return_tensors="pt",
|
89 |
+
).to("cuda")
|
90 |
|
|
|
91 |
generated = model.generate(**inputs, max_new_tokens=128)
|
92 |
+
trimmed = [
|
93 |
+
out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated)
|
94 |
+
]
|
95 |
+
raw_out = processor.batch_decode(
|
96 |
+
trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
|
97 |
+
)[0]
|
98 |
|
99 |
+
# βββββββ draw predicted click for quick visual verification (optional) ββββββ
|
100 |
try:
|
101 |
+
actions = ast.literal_eval(raw_out)
|
102 |
for act in actions if isinstance(actions, list) else [actions]:
|
103 |
+
pos = act.get("position")
|
104 |
+
if pos and isinstance(pos, list) and len(pos) == 2:
|
105 |
+
screenshot = draw_point(screenshot, pos)
|
106 |
+
except Exception:
|
107 |
+
# decoding failed β just return original screenshot
|
108 |
+
pass
|
109 |
+
|
110 |
+
return screenshot, raw_out
|
111 |
|
112 |
|
113 |
+
# ββββββββββββββββββββββββββ Gradio interface βββββββββββββββββββββββββββββββ
|
114 |
+
|
115 |
demo = gr.Interface(
|
116 |
fn=navigate,
|
117 |
inputs=[
|
118 |
gr.Image(type="pil", label="Screenshot"),
|
119 |
+
gr.Textbox(
|
120 |
+
lines=1,
|
121 |
+
placeholder="e.g. Search the weather for New York",
|
122 |
+
label="Task",
|
123 |
+
),
|
124 |
gr.Dropdown(choices=["web", "phone"], value="web", label="Platform"),
|
125 |
+
gr.JSON(label="Conversation History (list)", value=[]),
|
126 |
+
],
|
127 |
+
outputs=[
|
128 |
+
gr.Image(label="With Click Point"),
|
129 |
+
gr.Textbox(label="Raw Action JSON"),
|
130 |
],
|
131 |
+
title="ShowUIβ2B Navigation Demo",
|
|
|
132 |
)
|
133 |
|
134 |
demo.launch(
|
|
|
136 |
server_port=7860,
|
137 |
share=False, # or True if you need a public link
|
138 |
ssr_mode=False, # turn off experimental SSR so the process blocks
|
139 |
+
)
|