Spaces:
Sleeping
Sleeping
Delete audio.py
Browse files
audio.py
DELETED
@@ -1,58 +0,0 @@
|
|
1 |
-
import librosa
|
2 |
-
import numpy as np
|
3 |
-
|
4 |
-
def extract_features(audio_path):
|
5 |
-
y, sr = librosa.load(audio_path, sr=16000)
|
6 |
-
mfccs = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=13)
|
7 |
-
return np.mean(mfccs.T, axis=0)
|
8 |
-
|
9 |
-
# Example usage
|
10 |
-
features = extract_features("path/to/audio/file.wav")
|
11 |
-
|
12 |
-
from transformers import Wav2Vec2ForCTC, Wav2Vec2Tokenizer, MarianMTModel, MarianTokenizer
|
13 |
-
|
14 |
-
# Load pre-trained models
|
15 |
-
speech_recognition_model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-large-960h")
|
16 |
-
speech_recognition_tokenizer = Wav2Vec2Tokenizer.from_pretrained("facebook/wav2vec2-large-960h")
|
17 |
-
translation_model = MarianMTModel.from_pretrained("Helsinki-NLP/opus-mt-en-hi")
|
18 |
-
translation_tokenizer = MarianTokenizer.from_pretrained("Helsinki-NLP/opus-mt-en-hi")
|
19 |
-
|
20 |
-
from transformers import pipeline
|
21 |
-
|
22 |
-
# Example inference pipeline
|
23 |
-
def translate_audio(audio_path):
|
24 |
-
# Speech Recognition
|
25 |
-
speech_input = speech_recognition_tokenizer(extract_features(audio_path), return_tensors="pt").input_values
|
26 |
-
logits = speech_recognition_model(speech_input).logits
|
27 |
-
transcription = speech_recognition_tokenizer.batch_decode(torch.argmax(logits, dim=-1))[0]
|
28 |
-
|
29 |
-
# Translation
|
30 |
-
translated = translation_model.generate(**translation_tokenizer.prepare_seq2seq_batch(transcription, return_tensors="pt"))
|
31 |
-
translation = translation_tokenizer.batch_decode(translated, skip_special_tokens=True)[0]
|
32 |
-
|
33 |
-
return translation
|
34 |
-
|
35 |
-
# Save the models and tokenizer
|
36 |
-
speech_recognition_model.save_pretrained("path/to/save/wav2vec2")
|
37 |
-
speech_recognition_tokenizer.save_pretrained("path/to/save/wav2vec2")
|
38 |
-
translation_model.save_pretrained("path/to/save/opus-mt-en-hi")
|
39 |
-
translation_tokenizer.save_pretrained("path/to/save/opus-mt-en-hi")
|
40 |
-
|
41 |
-
# Upload to Hugging Face
|
42 |
-
!huggingface-cli login
|
43 |
-
!transformers-cli upload path/to/save/wav2vec2
|
44 |
-
!transformers-cli upload path/to/save/opus-mt-en-hi
|
45 |
-
|
46 |
-
from datetime import datetime
|
47 |
-
import pytz
|
48 |
-
|
49 |
-
def is_after_6_pm_ist():
|
50 |
-
ist = pytz.timezone('Asia/Kolkata')
|
51 |
-
current_time = datetime.now(ist)
|
52 |
-
return current_time.hour >= 18
|
53 |
-
|
54 |
-
if is_after_6_pm_ist():
|
55 |
-
translation = translate_audio("path/to/audio/file.wav")
|
56 |
-
print(translation)
|
57 |
-
else:
|
58 |
-
print("The translation service is available after 6 PM IST.")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|