Spaces:
Sleeping
Sleeping
Update audio_processing.py
Browse files- audio_processing.py +3 -3
audio_processing.py
CHANGED
@@ -55,7 +55,7 @@ class AudioProcessor:
|
|
55 |
"""Identify language of audio chunk"""
|
56 |
lid_model, lid_processor = models['lid']
|
57 |
inputs = lid_processor(audio_chunk, sampling_rate=16000, return_tensors="pt")
|
58 |
-
|
59 |
with torch.no_grad():
|
60 |
outputs = lid_model(inputs.input_values.to(self.device)).logits
|
61 |
lang_id = torch.argmax(outputs, dim=-1)[0].item()
|
@@ -70,7 +70,7 @@ class AudioProcessor:
|
|
70 |
|
71 |
mms_processor.tokenizer.set_target_lang(language)
|
72 |
mms_model.load_adapter(language)
|
73 |
-
|
74 |
inputs = mms_processor(audio_chunk, sampling_rate=16000, return_tensors="pt")
|
75 |
|
76 |
with torch.no_grad():
|
@@ -87,7 +87,7 @@ class AudioProcessor:
|
|
87 |
|
88 |
inputs = translation_tokenizer(text, return_tensors="pt")
|
89 |
inputs = inputs.to(self.device)
|
90 |
-
|
91 |
with torch.no_grad():
|
92 |
outputs = translation_model.generate(
|
93 |
**inputs,
|
|
|
55 |
"""Identify language of audio chunk"""
|
56 |
lid_model, lid_processor = models['lid']
|
57 |
inputs = lid_processor(audio_chunk, sampling_rate=16000, return_tensors="pt")
|
58 |
+
lid_model.to(self.device)
|
59 |
with torch.no_grad():
|
60 |
outputs = lid_model(inputs.input_values.to(self.device)).logits
|
61 |
lang_id = torch.argmax(outputs, dim=-1)[0].item()
|
|
|
70 |
|
71 |
mms_processor.tokenizer.set_target_lang(language)
|
72 |
mms_model.load_adapter(language)
|
73 |
+
mms_model.to(self.device)
|
74 |
inputs = mms_processor(audio_chunk, sampling_rate=16000, return_tensors="pt")
|
75 |
|
76 |
with torch.no_grad():
|
|
|
87 |
|
88 |
inputs = translation_tokenizer(text, return_tensors="pt")
|
89 |
inputs = inputs.to(self.device)
|
90 |
+
translation_model.to(self.device)
|
91 |
with torch.no_grad():
|
92 |
outputs = translation_model.generate(
|
93 |
**inputs,
|