Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -16,15 +16,14 @@ forced_decoder_ids = processor.get_decoder_prompt_ids(language="italian", task="
|
|
| 16 |
# Custom preprocessing function
|
| 17 |
def preprocess_audio(audio_data):
|
| 18 |
# Apply any custom preprocessing to the audio data here if needed
|
| 19 |
-
|
|
|
|
|
|
|
| 20 |
|
| 21 |
# Function to perform ASR on audio data
|
| 22 |
-
def transcribe_audio(
|
| 23 |
-
# Preprocess the audio data
|
| 24 |
-
input_features = preprocess_audio(audio_data)
|
| 25 |
-
|
| 26 |
# Generate token ids
|
| 27 |
-
predicted_ids = model.generate(input_features)
|
| 28 |
|
| 29 |
# Decode token ids to text
|
| 30 |
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)
|
|
|
|
| 16 |
# Custom preprocessing function
|
| 17 |
def preprocess_audio(audio_data):
|
| 18 |
# Apply any custom preprocessing to the audio data here if needed
|
| 19 |
+
# Ensure that the input data is a valid format for the model
|
| 20 |
+
processed_data = processor(audio_data, return_tensors="pt", padding=True, truncation=True)
|
| 21 |
+
return processed_data
|
| 22 |
|
| 23 |
# Function to perform ASR on audio data
|
| 24 |
+
def transcribe_audio(input_features):
|
|
|
|
|
|
|
|
|
|
| 25 |
# Generate token ids
|
| 26 |
+
predicted_ids = model.generate(**input_features)
|
| 27 |
|
| 28 |
# Decode token ids to text
|
| 29 |
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)
|