allandclive commited on
Commit
2df5a70
·
1 Parent(s): 22a0e2a

Update stitched_model.py

Browse files
Files changed (1) hide show
  1. stitched_model.py +1 -1
stitched_model.py CHANGED
@@ -16,7 +16,7 @@ class CombinedModel(nn.Module):
16
  # Use stt_model to transcribe the audio to text
17
  device = self.device
18
  audio = torch.tensor(batch["audio"][0]).to(self.device)
19
- input_features = self.stt_processor(audio,sampling_rate=16000, return_tensors="pt",max_length=300000, padding=True, truncation=True)
20
  stt_output = self.stt_model(input_features.input_values.to(device), attention_mask= input_features.attention_mask.to(device) )
21
  transcription = self.stt_processor.decode(torch.squeeze(stt_output.logits.argmax(axis=-1)).to(device))
22
  input_nmt_tokens = self.nmt_tokenizer(transcription, return_tensors="pt", padding=True, truncation=True)
 
16
  # Use stt_model to transcribe the audio to text
17
  device = self.device
18
  audio = torch.tensor(batch["audio"][0]).to(self.device)
19
+ input_features = self.stt_processor(audio,sampling_rate=16000, return_tensors="pt",max_length=220000, padding=True, truncation=True)
20
  stt_output = self.stt_model(input_features.input_values.to(device), attention_mask= input_features.attention_mask.to(device) )
21
  transcription = self.stt_processor.decode(torch.squeeze(stt_output.logits.argmax(axis=-1)).to(device))
22
  input_nmt_tokens = self.nmt_tokenizer(transcription, return_tensors="pt", padding=True, truncation=True)