Spaces:
Build error
Build error
patrickvonplaten
commited on
Commit
•
62a78c6
1
Parent(s):
63914a6
Update app.py
Browse files
app.py
CHANGED
@@ -1,29 +1,29 @@
|
|
1 |
-
import torch
|
2 |
-
from transformers import SpeechEncoderDecoder, Wav2Vec2Processor
|
3 |
import gradio as gr
|
4 |
-
import
|
|
|
|
|
|
|
|
|
5 |
|
6 |
-
def
|
7 |
-
|
8 |
-
|
9 |
-
|
10 |
-
|
11 |
-
|
|
|
12 |
|
13 |
-
def
|
14 |
-
|
15 |
-
input_values = processor(speech, sampling_rate=16_000, return_tensors="pt").input_values
|
16 |
-
logits = model(input_values).logits
|
17 |
-
predicted_ids = torch.argmax(logits, dim=-1)
|
18 |
-
transcription = processor.decode(predicted_ids[0], skip_special_tokens=True)
|
19 |
-
return transcription
|
20 |
|
|
|
21 |
|
22 |
-
|
23 |
-
|
24 |
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
|
|
|
|
|
|
|
1 |
import gradio as gr
|
2 |
+
import librosa
|
3 |
+
from transformers import Wav2Vec2Processor, SpeechEncoderDecoder
|
4 |
+
|
5 |
+
processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-xls-r-300m-en-to-15", use_auth_token="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")
|
6 |
+
model = SpeechEncoderDecoder.from_pretrained("facebook/wav2vec2-xls-r-300m-en-to-15", use_auth_token="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")
|
7 |
|
8 |
+
def process_audio_file(file):
|
9 |
+
data, sr = librosa.load(file)
|
10 |
+
if sr != 16000:
|
11 |
+
data = librosa.resample(data, sr, 16000)
|
12 |
+
print(data.shape)
|
13 |
+
input_values = processor(data, return_tensors="pt").input_values
|
14 |
+
return input_values
|
15 |
|
16 |
+
def transcribe(file):
|
17 |
+
input_values = process_audio_file(file)
|
|
|
|
|
|
|
|
|
|
|
18 |
|
19 |
+
sequences = model.generate(input_values, num_beams=1, max_length=30)
|
20 |
|
21 |
+
transcription = processor.batch_decode(sequences)
|
22 |
+
return transcription[0]
|
23 |
|
24 |
+
iface = gr.Interface(
|
25 |
+
fn=speech_recognize,
|
26 |
+
inputs=gr.inputs.Audio(source="microphone", type='filepath'),
|
27 |
+
outputs="text",
|
28 |
+
)
|
29 |
+
iface.launch()
|