LiamKhoaLe commited on
Commit
112f258
·
1 Parent(s): bd28ac3

Change model to turbo

Browse files
Files changed (2) hide show
  1. .DS_Store +0 -0
  2. app.py +6 -7
.DS_Store CHANGED
Binary files a/.DS_Store and b/.DS_Store differ
 
app.py CHANGED
@@ -14,18 +14,19 @@ from dotenv import load_dotenv
14
 
15
  load_dotenv()
16
 
17
- MODEL_NAME = "openai/whisper-large-v3"
18
  BATCH_SIZE = 8
19
- FILE_LIMIT_MB = 1000
20
- YT_LENGTH_LIMIT_S = 3600 # limit to 1 hour YouTube files
21
 
22
  device = 0 if torch.cuda.is_available() else "cpu"
23
 
24
  pipe = pipeline(
25
  task="automatic-speech-recognition",
26
  model=MODEL_NAME,
27
- chunk_length_s=30,
28
  device=device,
 
 
29
  )
30
 
31
 
@@ -110,9 +111,7 @@ mf_transcribe = gr.Interface(
110
  outputs=[gr.Textbox(label="Transcription"), gr.Textbox(label="Summary")],
111
  title="Whisper Large V3: Microphone",
112
  description=(
113
- "Transcribe long-form microphone or audio inputs with the click of a button! Demo uses the"
114
- f" checkpoint [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) and 🤗 Transformers to transcribe audio files"
115
- " of arbitrary length."
116
  ),
117
  allow_flagging="never",
118
  )
 
14
 
15
  load_dotenv()
16
 
17
+ MODEL_NAME = "openai/whisper-large-v3-turbo"
18
  BATCH_SIZE = 8
19
+ FILE_LIMIT_MB = 5000 # 5GB
20
+ YT_LENGTH_LIMIT_S = 7200 # 2 hours
21
 
22
  device = 0 if torch.cuda.is_available() else "cpu"
23
 
24
  pipe = pipeline(
25
  task="automatic-speech-recognition",
26
  model=MODEL_NAME,
 
27
  device=device,
28
+ ignore_warning=True,
29
+ model_kwargs={"torch_dtype": torch.float16} if torch.cuda.is_available() else {}
30
  )
31
 
32
 
 
111
  outputs=[gr.Textbox(label="Transcription"), gr.Textbox(label="Summary")],
112
  title="Whisper Large V3: Microphone",
113
  description=(
114
+ "Transcribe long-form microphone or audio inputs."
 
 
115
  ),
116
  allow_flagging="never",
117
  )