Spaces:
Running
on
Zero
Running
on
Zero
Darius Morawiec
commited on
Commit
·
3d8d21b
1
Parent(s):
bfa4ccc
Use constants for the default system prompt and examples
Browse files
app.py
CHANGED
|
@@ -45,6 +45,54 @@ MODEL_IDS = [
|
|
| 45 |
"Qwen/Qwen3-VL-8B-Instruct", # https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct
|
| 46 |
"Qwen/Qwen3-VL-32B-Instruct", # https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct
|
| 47 |
]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 48 |
|
| 49 |
# Global variables to track loaded model
|
| 50 |
current_model = None
|
|
@@ -111,11 +159,10 @@ with gr.Blocks() as demo:
|
|
| 111 |
choices=MODEL_IDS,
|
| 112 |
label="✨ Select Model ID",
|
| 113 |
)
|
| 114 |
-
default_system_prompt = 'You are a helpful assistant to detect objects in images. When asked to detect elements based on a description, you return a valid JSON object containing bounding boxes for all elements in the form `[{"bbox_2d": [xmin, ymin, xmax, ymax], "label": "placeholder"}, ...]`. For example, a valid response could be: `[{"bbox_2d": [10, 30, 20, 60], "label": "placeholder"}, {"bbox_2d": [40, 15, 52, 27], "label": "placeholder"}]`.'
|
| 115 |
system_prompt = gr.Textbox(
|
| 116 |
label="System Prompt",
|
| 117 |
lines=3,
|
| 118 |
-
value=
|
| 119 |
)
|
| 120 |
default_user_prompt = "detect object"
|
| 121 |
user_prompt = gr.Textbox(
|
|
@@ -320,54 +367,7 @@ with gr.Blocks() as demo:
|
|
| 320 |
cache_examples=True,
|
| 321 |
cache_mode="eager",
|
| 322 |
run_on_click=True,
|
| 323 |
-
examples=
|
| 324 |
-
[
|
| 325 |
-
EXAMPLES_DIR
|
| 326 |
-
/ "niklas-ohlrogge-niamoh-de-fDYRfHoRC4k-unsplash.jpg",
|
| 327 |
-
"Qwen/Qwen3-VL-4B-Instruct",
|
| 328 |
-
default_system_prompt,
|
| 329 |
-
"detect sailboat, rowboat, person",
|
| 330 |
-
512,
|
| 331 |
-
"Yes",
|
| 332 |
-
1920,
|
| 333 |
-
],
|
| 334 |
-
[
|
| 335 |
-
EXAMPLES_DIR / "elevate-nYgy58eb9aw-unsplash.jpg",
|
| 336 |
-
"Qwen/Qwen3-VL-4B-Instruct",
|
| 337 |
-
default_system_prompt,
|
| 338 |
-
"detect shirt, jeans, jacket, skirt, sunglasses, earring, drink",
|
| 339 |
-
1024,
|
| 340 |
-
"Yes",
|
| 341 |
-
1920,
|
| 342 |
-
],
|
| 343 |
-
[
|
| 344 |
-
EXAMPLES_DIR / "markus-spiske-oPDQGXW7i40-unsplash.jpg",
|
| 345 |
-
"Qwen/Qwen3-VL-4B-Instruct",
|
| 346 |
-
default_system_prompt,
|
| 347 |
-
"detect basketball, player with white jersey, player with black jersey",
|
| 348 |
-
512,
|
| 349 |
-
"Yes",
|
| 350 |
-
1920,
|
| 351 |
-
],
|
| 352 |
-
[
|
| 353 |
-
EXAMPLES_DIR / "william-hook-9e9PD9blAto-unsplash.jpg",
|
| 354 |
-
"Qwen/Qwen3-VL-4B-Instruct",
|
| 355 |
-
default_system_prompt,
|
| 356 |
-
"detect app to find great places, app to take beautiful photos, app to listen music",
|
| 357 |
-
512,
|
| 358 |
-
"Yes",
|
| 359 |
-
1920,
|
| 360 |
-
],
|
| 361 |
-
[
|
| 362 |
-
EXAMPLES_DIR / "tasso-mitsarakis-dw7Y4W6Rhmk-unsplash.jpg",
|
| 363 |
-
"Qwen/Qwen3-VL-4B-Instruct",
|
| 364 |
-
default_system_prompt,
|
| 365 |
-
"detect person, bicycle, netherlands flag",
|
| 366 |
-
1920,
|
| 367 |
-
"Yes",
|
| 368 |
-
1920,
|
| 369 |
-
],
|
| 370 |
-
],
|
| 371 |
inputs=[
|
| 372 |
image_input,
|
| 373 |
input_model_id,
|
|
|
|
| 45 |
"Qwen/Qwen3-VL-8B-Instruct", # https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct
|
| 46 |
"Qwen/Qwen3-VL-32B-Instruct", # https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct
|
| 47 |
]
|
| 48 |
+
DEFAULT_SYSTEM_PROMPT = 'You are a helpful assistant to detect objects in images. When asked to detect elements based on a description, you return a valid JSON object containing bounding boxes for all elements in the form `[{"bbox_2d": [xmin, ymin, xmax, ymax], "label": "placeholder"}, ...]`. For example, a valid response could be: `[{"bbox_2d": [10, 30, 20, 60], "label": "placeholder"}, {"bbox_2d": [40, 15, 52, 27], "label": "placeholder"}]`.'
|
| 49 |
+
EXAMPLES = [
|
| 50 |
+
[
|
| 51 |
+
EXAMPLES_DIR / "niklas-ohlrogge-niamoh-de-fDYRfHoRC4k-unsplash.jpg",
|
| 52 |
+
"Qwen/Qwen3-VL-4B-Instruct",
|
| 53 |
+
DEFAULT_SYSTEM_PROMPT,
|
| 54 |
+
"detect sailboat, rowboat, person",
|
| 55 |
+
512,
|
| 56 |
+
"Yes",
|
| 57 |
+
1920,
|
| 58 |
+
],
|
| 59 |
+
[
|
| 60 |
+
EXAMPLES_DIR / "elevate-nYgy58eb9aw-unsplash.jpg",
|
| 61 |
+
"Qwen/Qwen3-VL-4B-Instruct",
|
| 62 |
+
DEFAULT_SYSTEM_PROMPT,
|
| 63 |
+
"detect shirt, jeans, jacket, skirt, sunglasses, earring, drink",
|
| 64 |
+
1024,
|
| 65 |
+
"Yes",
|
| 66 |
+
1920,
|
| 67 |
+
],
|
| 68 |
+
[
|
| 69 |
+
EXAMPLES_DIR / "markus-spiske-oPDQGXW7i40-unsplash.jpg",
|
| 70 |
+
"Qwen/Qwen3-VL-4B-Instruct",
|
| 71 |
+
DEFAULT_SYSTEM_PROMPT,
|
| 72 |
+
"detect basketball, player with white jersey, player with black jersey",
|
| 73 |
+
512,
|
| 74 |
+
"Yes",
|
| 75 |
+
1920,
|
| 76 |
+
],
|
| 77 |
+
[
|
| 78 |
+
EXAMPLES_DIR / "william-hook-9e9PD9blAto-unsplash.jpg",
|
| 79 |
+
"Qwen/Qwen3-VL-4B-Instruct",
|
| 80 |
+
DEFAULT_SYSTEM_PROMPT,
|
| 81 |
+
"detect app to find great places, app to take beautiful photos, app to listen music",
|
| 82 |
+
512,
|
| 83 |
+
"Yes",
|
| 84 |
+
1920,
|
| 85 |
+
],
|
| 86 |
+
[
|
| 87 |
+
EXAMPLES_DIR / "tasso-mitsarakis-dw7Y4W6Rhmk-unsplash.jpg",
|
| 88 |
+
"Qwen/Qwen3-VL-4B-Instruct",
|
| 89 |
+
DEFAULT_SYSTEM_PROMPT,
|
| 90 |
+
"detect person, bicycle, netherlands flag",
|
| 91 |
+
1920,
|
| 92 |
+
"Yes",
|
| 93 |
+
1920,
|
| 94 |
+
],
|
| 95 |
+
]
|
| 96 |
|
| 97 |
# Global variables to track loaded model
|
| 98 |
current_model = None
|
|
|
|
| 159 |
choices=MODEL_IDS,
|
| 160 |
label="✨ Select Model ID",
|
| 161 |
)
|
|
|
|
| 162 |
system_prompt = gr.Textbox(
|
| 163 |
label="System Prompt",
|
| 164 |
lines=3,
|
| 165 |
+
value=DEFAULT_SYSTEM_PROMPT,
|
| 166 |
)
|
| 167 |
default_user_prompt = "detect object"
|
| 168 |
user_prompt = gr.Textbox(
|
|
|
|
| 367 |
cache_examples=True,
|
| 368 |
cache_mode="eager",
|
| 369 |
run_on_click=True,
|
| 370 |
+
examples=EXAMPLES,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 371 |
inputs=[
|
| 372 |
image_input,
|
| 373 |
input_model_id,
|