prithivMLmods commited on
Commit
2bb8d53
·
verified ·
1 Parent(s): e62211b

update app

Browse files
Files changed (1) hide show
  1. app.py +16 -16
app.py CHANGED
@@ -51,7 +51,7 @@ class ThistleTheme(Soft):
51
  neutral_hue: colors.Color | str = colors.slate,
52
  text_size: sizes.Size | str = sizes.text_lg,
53
  font: fonts.Font | str | Iterable[fonts.Font | str] = (
54
- fonts.GoogleFont("Inconsolata"), "Arial", "sans-serif",
55
  ),
56
  font_mono: fonts.Font | str | Iterable[fonts.Font | str] = (
57
  fonts.GoogleFont("IBM Plex Mono"), "ui-monospace", "monospace",
@@ -372,17 +372,17 @@ def generate_gif(text: str, gif_path: str, max_new_tokens: int = 1024, temperatu
372
  buffer = buffer.replace("<|im_end|>", "")
373
  time.sleep(0.01)
374
  yield buffer, buffer
375
-
376
- image_examples = [["Perform OCR on the image precisely and reconstruct it correctly...", "examples/images/1.jpg"],
377
  ["Caption the image. Describe the safety measures shown in the image. Conclude whether the situation is (safe or unsafe)...", "examples/images/2.jpg"],
378
  ["Solve the problem...", "examples/images/3.png"]]
379
- video_examples = [["Explain the Ad video in detail.", "examples/videos/1.mp4"],
380
  ["Explain the video in detail.", "examples/videos/2.mp4"]]
381
- pdf_examples = [["Extract the content precisely.", "examples/pdfs/doc1.pdf"],
382
  ["Analyze and provide a short report.", "examples/pdfs/doc2.pdf"]]
383
  gif_examples = [["Describe this GIF.", "examples/gifs/1.gif"],
384
  ["Describe this GIF.", "examples/gifs/2.gif"]]
385
- caption_examples = [["https://huggingface.co/datasets/merve/vlm_test_images/resolve/main/candy.JPG"],
386
  ["examples/captions/2.png"], ["examples/captions/3.png"]]
387
 
388
  with gr.Blocks(theme=thistle_theme, css=css) as demo:
@@ -440,21 +440,21 @@ with gr.Blocks(theme=thistle_theme, css=css) as demo:
440
  output = gr.Textbox(label="Raw Output Stream", interactive=False, lines=14, show_copy_button=True)
441
  with gr.Accordion("(Result.md)", open=False):
442
  markdown_output = gr.Markdown(label="(Result.Md)")
443
-
444
- image_submit.click(fn=generate_image,
445
- inputs=[image_query, image_upload, max_new_tokens, temperature, top_p, top_k, repetition_penalty],
446
  outputs=[output, markdown_output])
447
- video_submit.click(fn=generate_video,
448
- inputs=[video_query, video_upload, max_new_tokens, temperature, top_p, top_k, repetition_penalty],
449
  outputs=[output, markdown_output])
450
  pdf_submit.click(fn=generate_pdf,
451
- inputs=[pdf_query, pdf_state, max_new_tokens, temperature, top_p, top_k, repetition_penalty],
452
  outputs=[output, markdown_output])
453
- gif_submit.click(fn=generate_gif,
454
- inputs=[gif_query, gif_upload, max_new_tokens, temperature, top_p, top_k, repetition_penalty],
455
  outputs=[output, markdown_output])
456
- caption_submit.click(fn=generate_caption,
457
- inputs=[caption_image_upload, max_new_tokens, temperature, top_p, top_k, repetition_penalty],
458
  outputs=[output, markdown_output])
459
 
460
  pdf_upload.change(fn=load_and_preview_pdf, inputs=[pdf_upload], outputs=[pdf_preview_img, pdf_state, page_info])
 
51
  neutral_hue: colors.Color | str = colors.slate,
52
  text_size: sizes.Size | str = sizes.text_lg,
53
  font: fonts.Font | str | Iterable[fonts.Font | str] = (
54
+ fonts.GoogleFont("Outfit"), "Arial", "sans-serif",
55
  ),
56
  font_mono: fonts.Font | str | Iterable[fonts.Font | str] = (
57
  fonts.GoogleFont("IBM Plex Mono"), "ui-monospace", "monospace",
 
372
  buffer = buffer.replace("<|im_end|>", "")
373
  time.sleep(0.01)
374
  yield buffer, buffer
375
+
376
+ image_examples = [["Perform OCR on the image precisely and reconstruct it correctly...", "examples/images/1.jpg"],
377
  ["Caption the image. Describe the safety measures shown in the image. Conclude whether the situation is (safe or unsafe)...", "examples/images/2.jpg"],
378
  ["Solve the problem...", "examples/images/3.png"]]
379
+ video_examples = [["Explain the Ad video in detail.", "examples/videos/1.mp4"],
380
  ["Explain the video in detail.", "examples/videos/2.mp4"]]
381
+ pdf_examples = [["Extract the content precisely.", "examples/pdfs/doc1.pdf"],
382
  ["Analyze and provide a short report.", "examples/pdfs/doc2.pdf"]]
383
  gif_examples = [["Describe this GIF.", "examples/gifs/1.gif"],
384
  ["Describe this GIF.", "examples/gifs/2.gif"]]
385
+ caption_examples = [["https://huggingface.co/datasets/merve/vlm_test_images/resolve/main/candy.JPG"],
386
  ["examples/captions/2.png"], ["examples/captions/3.png"]]
387
 
388
  with gr.Blocks(theme=thistle_theme, css=css) as demo:
 
440
  output = gr.Textbox(label="Raw Output Stream", interactive=False, lines=14, show_copy_button=True)
441
  with gr.Accordion("(Result.md)", open=False):
442
  markdown_output = gr.Markdown(label="(Result.Md)")
443
+
444
+ image_submit.click(fn=generate_image,
445
+ inputs=[image_query, image_upload, max_new_tokens, temperature, top_p, top_k, repetition_penalty],
446
  outputs=[output, markdown_output])
447
+ video_submit.click(fn=generate_video,
448
+ inputs=[video_query, video_upload, max_new_tokens, temperature, top_p, top_k, repetition_penalty],
449
  outputs=[output, markdown_output])
450
  pdf_submit.click(fn=generate_pdf,
451
+ inputs=[pdf_query, pdf_state, max_new_tokens, temperature, top_p, top_k, repetition_penalty],
452
  outputs=[output, markdown_output])
453
+ gif_submit.click(fn=generate_gif,
454
+ inputs=[gif_query, gif_upload, max_new_tokens, temperature, top_p, top_k, repetition_penalty],
455
  outputs=[output, markdown_output])
456
+ caption_submit.click(fn=generate_caption,
457
+ inputs=[caption_image_upload, max_new_tokens, temperature, top_p, top_k, repetition_penalty],
458
  outputs=[output, markdown_output])
459
 
460
  pdf_upload.change(fn=load_and_preview_pdf, inputs=[pdf_upload], outputs=[pdf_preview_img, pdf_state, page_info])