Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -27,7 +27,6 @@ from docling_core.types.doc import DoclingDocument, DocTagsDocument
|
|
27 |
import re
|
28 |
import ast
|
29 |
import html
|
30 |
-
import urllib.parse
|
31 |
|
32 |
# Constants for text generation
|
33 |
MAX_MAX_NEW_TOKENS = 2048
|
@@ -226,7 +225,7 @@ def generate_video(model_name: str, text: str, video_path: str,
|
|
226 |
yield "Invalid model selected.", "Invalid model selected."
|
227 |
return
|
228 |
|
229 |
-
if video_path is None:
|
230 |
yield "Please upload a video.", "Please upload a video."
|
231 |
return
|
232 |
|
@@ -298,84 +297,99 @@ video_examples = [
|
|
298 |
["Explain the video in detail.", "videos/2.mp4"]
|
299 |
]
|
300 |
|
301 |
-
#
|
302 |
-
|
303 |
-
|
304 |
-
|
305 |
-
|
306 |
-
|
307 |
-
|
308 |
-
|
309 |
-
# Updated CSS with fancy-button styles
|
310 |
-
css = f"""
|
311 |
-
.fancy-button {{
|
312 |
-
--round: 0.75rem;
|
313 |
cursor: pointer;
|
314 |
position: relative;
|
315 |
-
|
316 |
-
|
317 |
-
|
318 |
-
|
319 |
-
|
320 |
-
|
321 |
-
|
322 |
-
|
323 |
-
|
324 |
-
|
325 |
-
|
326 |
-
|
327 |
-
|
328 |
-
|
329 |
-
|
330 |
-
|
331 |
-
|
332 |
-
|
333 |
-
|
334 |
-
|
335 |
-
|
336 |
-
|
337 |
-
|
338 |
-
|
339 |
-
|
340 |
-
|
341 |
-
|
342 |
-
|
343 |
-
|
344 |
-
|
345 |
-
}
|
346 |
-
|
347 |
-
.
|
348 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
349 |
position: absolute;
|
350 |
-
|
351 |
-
|
352 |
-
|
353 |
-
|
354 |
-
|
355 |
-
|
356 |
-
|
357 |
-
|
358 |
-
|
359 |
-
|
360 |
-
|
361 |
-
|
362 |
-
|
363 |
-
|
364 |
-
|
365 |
-
|
366 |
-
|
367 |
-
|
368 |
-
|
369 |
-
|
370 |
-
|
371 |
-
|
372 |
-
|
373 |
-
|
374 |
-
|
375 |
-
|
376 |
-
|
377 |
-
|
378 |
-
|
|
|
|
|
|
|
|
|
|
|
379 |
"""
|
380 |
|
381 |
# Create the Gradio Interface
|
@@ -387,7 +401,7 @@ with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
|
|
387 |
with gr.TabItem("Image Inference"):
|
388 |
image_query = gr.Textbox(label="Query Input", placeholder="Enter your query here...")
|
389 |
image_upload = gr.Image(type="pil", label="Image")
|
390 |
-
|
391 |
gr.Examples(
|
392 |
examples=image_examples,
|
393 |
inputs=[image_query, image_upload]
|
@@ -395,7 +409,7 @@ with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
|
|
395 |
with gr.TabItem("Video Inference"):
|
396 |
video_query = gr.Textbox(label="Query Input", placeholder="Enter your query here...")
|
397 |
video_upload = gr.Video(label="Video")
|
398 |
-
video_submit = gr.Button("Submit", elem_classes="
|
399 |
gr.Examples(
|
400 |
examples=video_examples,
|
401 |
inputs=[video_query, video_upload]
|
@@ -411,7 +425,6 @@ with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
|
|
411 |
with gr.Column(elem_classes="canvas-output"):
|
412 |
gr.Markdown("## Output")
|
413 |
raw_output = gr.Textbox(label="Raw Output Stream", interactive=False, lines=2)
|
414 |
-
|
415 |
with gr.Accordion("(Result.md)", open=False):
|
416 |
formatted_output = gr.Markdown(label="(Result.md)")
|
417 |
|
|
|
27 |
import re
|
28 |
import ast
|
29 |
import html
|
|
|
30 |
|
31 |
# Constants for text generation
|
32 |
MAX_MAX_NEW_TOKENS = 2048
|
|
|
225 |
yield "Invalid model selected.", "Invalid model selected."
|
226 |
return
|
227 |
|
228 |
+
Nm if video_path is None:
|
229 |
yield "Please upload a video.", "Please upload a video."
|
230 |
return
|
231 |
|
|
|
297 |
["Explain the video in detail.", "videos/2.mp4"]
|
298 |
]
|
299 |
|
300 |
+
# Updated CSS with the new Submit button theme
|
301 |
+
css = """
|
302 |
+
.submit-btn {
|
303 |
+
--stone-50: #fafaf9;
|
304 |
+
--stone-800: #292524;
|
305 |
+
--yellow-400: #facc15;
|
306 |
+
|
307 |
+
font-size: 1rem;
|
|
|
|
|
|
|
|
|
308 |
cursor: pointer;
|
309 |
position: relative;
|
310 |
+
font-family: "Rubik", sans-serif;
|
311 |
+
font-weight: bold;
|
312 |
+
line-height: 1;
|
313 |
+
padding: 0.75rem 1.25rem;
|
314 |
+
transform: translate(-4px, -4px);
|
315 |
+
outline: 2px solid transparent;
|
316 |
+
outline-offset: 5px;
|
317 |
+
border-radius: 9999px;
|
318 |
+
background-color: var(--yellow-400);
|
319 |
+
color: var(--stone-800);
|
320 |
+
transition:
|
321 |
+
transform 150ms ease,
|
322 |
+
box-shadow 150ms ease;
|
323 |
+
text-align: center;
|
324 |
+
box-shadow:
|
325 |
+
0.5px 0.5px 0 0 var(--stone-800),
|
326 |
+
1px 1px 0 0 var(--stone-800),
|
327 |
+
1.5px 1.5px 0 0 var(--stone-800),
|
328 |
+
2px 2px 0 0 var(--stone-800),
|
329 |
+
2.5px 2.5px 0 0 var(--stone-800),
|
330 |
+
3px 3px 0 0 var(--stone-800),
|
331 |
+
0 0 0 2px var(--stone-50),
|
332 |
+
0.5px 0.5px 0 2px var(--stone-50),
|
333 |
+
1px 1px 0 2px var(--stone-50),
|
334 |
+
1.5px 1.5px 0 2px var(--stone-50),
|
335 |
+
2px 2px 0 2px var(--stone-50),
|
336 |
+
2.5px 2.5px 0 2px var(--stone-50),
|
337 |
+
3px 3px 0 2px var(--stone-50),
|
338 |
+
3.5px 3.5px 0 2px var(--stone-50),
|
339 |
+
4px 4px 0 2px var(--stone-50);
|
340 |
+
}
|
341 |
+
|
342 |
+
.submit-btn:hover {
|
343 |
+
transform: translate(0, 0);
|
344 |
+
box-shadow: 0 0 0 2px var(--stone-50);
|
345 |
+
}
|
346 |
+
|
347 |
+
.submit-btn:active {
|
348 |
+
transform: translate(0, 2px);
|
349 |
+
}
|
350 |
+
|
351 |
+
.submit-btn:focus-visible {
|
352 |
+
outline-color: var(--yellow-400);
|
353 |
+
outline-style: dashed;
|
354 |
+
}
|
355 |
+
|
356 |
+
.submit-btn::before {
|
357 |
+
content: "";
|
358 |
position: absolute;
|
359 |
+
inset: 0;
|
360 |
+
border-radius: 9999px;
|
361 |
+
opacity: 0.5;
|
362 |
+
background-image: radial-gradient(
|
363 |
+
rgb(255 255 255 / 80%) 20%,
|
364 |
+
transparent 20%
|
365 |
+
),
|
366 |
+
radial-gradient(rgb(255 255 255 / 100%) 20%, transparent 20%);
|
367 |
+
background-position:
|
368 |
+
0 0,
|
369 |
+
4px 4px;
|
370 |
+
background-size: 8px 8px;
|
371 |
+
mix-blend-mode: hard-light;
|
372 |
+
animation: dots 0.5s infinite linear;
|
373 |
+
}
|
374 |
+
|
375 |
+
@keyframes dots {
|
376 |
+
0% {
|
377 |
+
background-position:
|
378 |
+
0 0,
|
379 |
+
4px 4px;
|
380 |
+
}
|
381 |
+
100% {
|
382 |
+
background-position:
|
383 |
+
8px 0,
|
384 |
+
12px 4px;
|
385 |
+
}
|
386 |
+
}
|
387 |
+
|
388 |
+
.canvas-output {
|
389 |
+
border: 2px solid #4682B4;
|
390 |
+
border-radius: 10px;
|
391 |
+
padding: 20px;
|
392 |
+
}
|
393 |
"""
|
394 |
|
395 |
# Create the Gradio Interface
|
|
|
401 |
with gr.TabItem("Image Inference"):
|
402 |
image_query = gr.Textbox(label="Query Input", placeholder="Enter your query here...")
|
403 |
image_upload = gr.Image(type="pil", label="Image")
|
404 |
+
image Timage_submit = gr.Button("Submit", elem_classes="submit-btn")
|
405 |
gr.Examples(
|
406 |
examples=image_examples,
|
407 |
inputs=[image_query, image_upload]
|
|
|
409 |
with gr.TabItem("Video Inference"):
|
410 |
video_query = gr.Textbox(label="Query Input", placeholder="Enter your query here...")
|
411 |
video_upload = gr.Video(label="Video")
|
412 |
+
video_submit = gr.Button("Submit", elem_classes="submit-btn")
|
413 |
gr.Examples(
|
414 |
examples=video_examples,
|
415 |
inputs=[video_query, video_upload]
|
|
|
425 |
with gr.Column(elem_classes="canvas-output"):
|
426 |
gr.Markdown("## Output")
|
427 |
raw_output = gr.Textbox(label="Raw Output Stream", interactive=False, lines=2)
|
|
|
428 |
with gr.Accordion("(Result.md)", open=False):
|
429 |
formatted_output = gr.Markdown(label="(Result.md)")
|
430 |
|