WJ88 commited on
Commit
d8eb8af
·
verified ·
1 Parent(s): 65c1721

i hate ui v5657

Browse files
Files changed (1) hide show
  1. app.py +27 -23
app.py CHANGED
@@ -291,13 +291,35 @@ with gr.Blocks(title="Parakeet-TDT v3 (Unified MALSD Beam)") as demo:
291
  # Header
292
  gr.Markdown("### RELEASE: GIGA-CHAD-v.0.7")
293
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
294
  # Interactive Tabs
295
  with gr.Tab("Mic"):
296
  # Live audio input and output components
297
  mic = gr.Audio(
298
  sources=["microphone"], type="numpy", streaming=True, label="Speak"
299
  )
300
- text_out = gr.Textbox(label="Transcript", lines=8)
301
  flush_btn = gr.Button("Flush")
302
 
303
  # Session state for maintaining context across streams
@@ -319,8 +341,9 @@ with gr.Blocks(title="Parakeet-TDT v3 (Unified MALSD Beam)") as demo:
319
  results_table = gr.Dataframe(
320
  headers=["file", "text"],
321
  label="Results",
322
- row_count=(0, "dynamic"),
323
  col_count=(2, "fixed"),
 
324
  )
325
  # Event: Process uploaded files and update table
326
  run_btn.click(files_run, inputs=[files], outputs=[results_table])
@@ -393,27 +416,6 @@ with gr.Blocks(title="Parakeet-TDT v3 (Unified MALSD Beam)") as demo:
393
  )
394
  gr.HTML(highlights_html)
395
 
396
- # Features Table (Static Data)
397
- features_data = [
398
- ["Model Setup", "Loads Parakeet-TDT-0.6b-v3 (RNNT-based) with MALSD "
399
- "decoding for beam exploration and loop labels for alignments."],
400
- ["Audio Handling", "Resamples to 16kHz mono, supports various formats."],
401
- ["Streaming (Mic)", "Partial hypotheses for seamless updates, "
402
- "session-based for multi-chunk context."],
403
- ["UI", "Gradio tabs—Mic for live input/output (flush to finalize), "
404
- "Files for batch results table."],
405
- ["Tech Stack", "NeMo (ASR core), Gradio (web UI), Torchaudio/Soundfile "
406
- "(audio utils)."],
407
- ]
408
- gr.Dataframe(
409
- value=features_data,
410
- headers=["Feature", "Description"],
411
- datatype=["text", "text"],
412
- row_count=(len(features_data), "fixed"),
413
- col_count=(2, "fixed"),
414
- interactive=False,
415
- )
416
-
417
  # TODO Section
418
  todo_html = (
419
  "<h3>TODO:</h3>"
@@ -426,4 +428,6 @@ with gr.Blocks(title="Parakeet-TDT v3 (Unified MALSD Beam)") as demo:
426
  )
427
  gr.HTML(todo_html)
428
 
 
 
429
  demo.queue().launch(ssr_mode=False)
 
291
  # Header
292
  gr.Markdown("### RELEASE: GIGA-CHAD-v.0.7")
293
 
294
+ # Features Table (Static Data) - Moved here for prominence after title
295
+ features_data = [
296
+ ["Model Setup", "Loads Parakeet-TDT-0.6b-v3 (RNNT-based) with MALSD "
297
+ "decoding for beam exploration and loop labels for alignments."],
298
+ ["Audio Handling", "Resamples to 16kHz mono, supports various formats."],
299
+ ["Streaming (Mic)", "Partial hypotheses for seamless updates, "
300
+ "session-based for multi-chunk context."],
301
+ ["UI", "Gradio tabs—Mic for live input/output (flush to finalize), "
302
+ "Files for batch results table."],
303
+ ["Tech Stack", "NeMo (ASR core), Gradio (web UI), Torchaudio/Soundfile "
304
+ "(audio utils)."],
305
+ ]
306
+ gr.Dataframe(
307
+ value=features_data,
308
+ headers=["Feature", "Description"],
309
+ datatype=["text", "text"],
310
+ row_count=(len(features_data), "fixed"),
311
+ col_count=(2, "fixed"),
312
+ interactive=False,
313
+ wrap=True, # Ensures text wraps for better visibility in cells
314
+ )
315
+
316
  # Interactive Tabs
317
  with gr.Tab("Mic"):
318
  # Live audio input and output components
319
  mic = gr.Audio(
320
  sources=["microphone"], type="numpy", streaming=True, label="Speak"
321
  )
322
+ text_out = gr.Textbox(label="Transcript", lines=4) # Reduced height for visibility
323
  flush_btn = gr.Button("Flush")
324
 
325
  # Session state for maintaining context across streams
 
341
  results_table = gr.Dataframe(
342
  headers=["file", "text"],
343
  label="Results",
344
+ row_count=(5, "dynamic"), # Limited initial rows for smaller window
345
  col_count=(2, "fixed"),
346
+ wrap=True, # Text wraps to keep visible without horizontal scroll
347
  )
348
  # Event: Process uploaded files and update table
349
  run_btn.click(files_run, inputs=[files], outputs=[results_table])
 
416
  )
417
  gr.HTML(highlights_html)
418
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
419
  # TODO Section
420
  todo_html = (
421
  "<h3>TODO:</h3>"
 
428
  )
429
  gr.HTML(todo_html)
430
 
431
+ # Launch Configuration
432
+ # Note: ssr_mode=False disables server-side rendering for better real-time streaming performance.
433
  demo.queue().launch(ssr_mode=False)