leo-kwan commited on
Commit
052ba4a
1 Parent(s): 5b43f39

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +2 -2
app.py CHANGED
@@ -9,7 +9,7 @@ from transformers import SpeechT5ForTextToSpeech, SpeechT5HifiGan, SpeechT5Proce
9
  device = "cuda:0" if torch.cuda.is_available() else "cpu"
10
 
11
  # load speech translation checkpoint
12
- asr_pipe = pipeline("automatic-speech-recognition", model="openai/whisper-base", device=device)
13
 
14
  # load text-to-speech checkpoint and speaker embeddings
15
  processor = SpeechT5Processor.from_pretrained("leo-kwan/speecht5_finetuned_voxpopuli_lt")
@@ -22,7 +22,7 @@ speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze
22
 
23
 
24
  def translate(audio):
25
- outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "translate", "language": "french"})
26
  return outputs["text"]
27
 
28
 
 
9
  device = "cuda:0" if torch.cuda.is_available() else "cpu"
10
 
11
  # load speech translation checkpoint
12
+ asr_pipe = pipeline("automatic-speech-recognition", model="openai/whisper-base", language = 'french', device=device)
13
 
14
  # load text-to-speech checkpoint and speaker embeddings
15
  processor = SpeechT5Processor.from_pretrained("leo-kwan/speecht5_finetuned_voxpopuli_lt")
 
22
 
23
 
24
  def translate(audio):
25
+ outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "translate"})
26
  return outputs["text"]
27
 
28