Spaces:
Build error
Build error
patrickvonplaten
commited on
Commit
•
9900ba7
1
Parent(s):
04e8018
Update app.py
Browse files
app.py
CHANGED
@@ -1,18 +1,8 @@
|
|
1 |
-
import soundfile as sf
|
2 |
import torch
|
3 |
from transformers import SpeechEncoderDecoder, Wav2Vec2Processor
|
4 |
import gradio as gr
|
5 |
import scipy.signal as sps
|
6 |
-
import sox
|
7 |
|
8 |
-
def convert(inputfile, outfile):
|
9 |
-
sox_tfm = sox.Transformer()
|
10 |
-
sox_tfm.set_output_format(
|
11 |
-
file_type="wav", channels=1, encoding="signed-integer", rate=16000, bits=16
|
12 |
-
)
|
13 |
-
#print(this is not done)
|
14 |
-
sox_tfm.build(inputfile, outfile)
|
15 |
-
|
16 |
def read_file(wav):
|
17 |
sample_rate, signal = wav
|
18 |
signal = signal.mean(-1)
|
@@ -21,12 +11,6 @@ def read_file(wav):
|
|
21 |
return resampled_signal
|
22 |
|
23 |
def parse_transcription(wav_file):
|
24 |
-
'''
|
25 |
-
filename = wav_file.name.split('.')[0]
|
26 |
-
convert(wav_file.name, filename + "16k.wav")
|
27 |
-
speech, _ = sf.read(filename + "16k.wav")
|
28 |
-
|
29 |
-
'''
|
30 |
speech = read_file(wav_file)
|
31 |
input_values = processor(speech, sampling_rate=16_000, return_tensors="pt").input_values
|
32 |
logits = model(input_values).logits
|
@@ -35,14 +19,10 @@ def parse_transcription(wav_file):
|
|
35 |
return transcription
|
36 |
|
37 |
|
38 |
-
processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-xls-r-300m-en-to-15")
|
39 |
-
model = SpeechEncoderDecoder.from_pretrained("facebook/wav2vec2-xls-r-300m-en-to-15")
|
40 |
-
|
41 |
-
|
42 |
-
processor = Wav2Vec2Processor.from_pretrained("Harveenchadha/vakyansh-wav2vec2-hindi-him-4200")
|
43 |
-
model = Wav2Vec2ForCTC.from_pretrained("Harveenchadha/vakyansh-wav2vec2-hindi-him-4200")
|
44 |
-
|
45 |
|
|
|
46 |
#input_ = gr.inputs.Audio(source="microphone", type="file")
|
47 |
input_ = gr.inputs.Audio(source="microphone", type="numpy")
|
48 |
gr.Interface(parse_transcription, inputs = input_, outputs="text",
|
|
|
|
|
1 |
import torch
|
2 |
from transformers import SpeechEncoderDecoder, Wav2Vec2Processor
|
3 |
import gradio as gr
|
4 |
import scipy.signal as sps
|
|
|
5 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
6 |
def read_file(wav):
|
7 |
sample_rate, signal = wav
|
8 |
signal = signal.mean(-1)
|
|
|
11 |
return resampled_signal
|
12 |
|
13 |
def parse_transcription(wav_file):
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
speech = read_file(wav_file)
|
15 |
input_values = processor(speech, sampling_rate=16_000, return_tensors="pt").input_values
|
16 |
logits = model(input_values).logits
|
|
|
19 |
return transcription
|
20 |
|
21 |
|
22 |
+
processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-xls-r-300m-en-to-15", use_auth_token="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")
|
23 |
+
model = SpeechEncoderDecoder.from_pretrained("facebook/wav2vec2-xls-r-300m-en-to-15", use_auth_token="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")
|
|
|
|
|
|
|
|
|
|
|
24 |
|
25 |
+
|
26 |
#input_ = gr.inputs.Audio(source="microphone", type="file")
|
27 |
input_ = gr.inputs.Audio(source="microphone", type="numpy")
|
28 |
gr.Interface(parse_transcription, inputs = input_, outputs="text",
|