jpdiazpardo commited on
Commit
b8b135b
1 Parent(s): 2809c68

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +2 -23
app.py CHANGED
@@ -1,6 +1,7 @@
 
1
  import torch
2
  from transformers import pipeline
3
- import gradio as gr
4
 
5
  MODEL_NAME = "openai/whisper-medium"
6
  BATCH_SIZE = 8
@@ -14,28 +15,6 @@ pipe = pipeline(
14
  device=device,
15
  )
16
 
17
-
18
- # Copied from https://github.com/openai/whisper/blob/c09a7ae299c4c34c5839a76380ae407e7d785914/whisper/utils.py#L50
19
- def format_timestamp(seconds: float, always_include_hours: bool = False, decimal_marker: str = "."):
20
- if seconds is not None:
21
- milliseconds = round(seconds * 1000.0)
22
-
23
- hours = milliseconds // 3_600_000
24
- milliseconds -= hours * 3_600_000
25
-
26
- minutes = milliseconds // 60_000
27
- milliseconds -= minutes * 60_000
28
-
29
- seconds = milliseconds // 1_000
30
- milliseconds -= seconds * 1_000
31
-
32
- hours_marker = f"{hours:02d}:" if always_include_hours or hours > 0 else ""
33
- return f"{hours_marker}{minutes:02d}:{seconds:02d}{decimal_marker}{milliseconds:03d}"
34
- else:
35
- # we have a malformed timestamp so just return it as is
36
- return seconds
37
-
38
-
39
  def transcribe(file, task, return_timestamps):
40
  outputs = pipe(file, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=return_timestamps)
41
  text = outputs["text"]
 
1
+ import gradio as gr
2
  import torch
3
  from transformers import pipeline
4
+ from timestamp import format_timestamp
5
 
6
  MODEL_NAME = "openai/whisper-medium"
7
  BATCH_SIZE = 8
 
15
  device=device,
16
  )
17
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
18
  def transcribe(file, task, return_timestamps):
19
  outputs = pipe(file, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=return_timestamps)
20
  text = outputs["text"]