Update README.md
Browse files
README.md
CHANGED
@@ -13,44 +13,44 @@ library_name: transformers
|
|
13 |
---
|
14 |
how to use the model in colab:
|
15 |
|
16 |
-
#start
|
17 |
-
pip install torch torchaudio transformers librosa gradio
|
18 |
-
from transformers import WhisperProcessor, WhisperForConditionalGeneration
|
19 |
-
import torch
|
20 |
-
|
21 |
-
#Load your fine-tuned Whisper model and processor
|
22 |
-
model_name = "hackergeek98/tinyyyy_whisper"
|
23 |
-
processor = WhisperProcessor.from_pretrained(model_name)
|
24 |
-
model = WhisperForConditionalGeneration.from_pretrained(model_name)
|
25 |
-
|
26 |
-
#Force the model to transcribe in Persian
|
27 |
-
model.config.forced_decoder_ids = processor.get_decoder_prompt_ids(language="fa", task="transcribe")
|
28 |
-
|
29 |
-
#Move model to GPU if available
|
30 |
-
device = "cuda" if torch.cuda.is_available() else "cpu"
|
31 |
-
model.to(device)
|
32 |
-
import librosa
|
33 |
-
|
34 |
-
def transcribe_audio(audio_file):
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
from google.colab import files
|
49 |
-
|
50 |
-
#Upload an audio file
|
51 |
-
uploaded = files.upload()
|
52 |
-
audio_file = list(uploaded.keys())[0]
|
53 |
-
|
54 |
-
#Transcribe the audio
|
55 |
-
transcription = transcribe_audio(audio_file)
|
56 |
-
print("Transcription:", transcription)
|
|
|
13 |
---
|
14 |
how to use the model in colab:
|
15 |
|
16 |
+
#start
|
17 |
+
pip install torch torchaudio transformers librosa gradio
|
18 |
+
from transformers import WhisperProcessor, WhisperForConditionalGeneration
|
19 |
+
import torch
|
20 |
+
|
21 |
+
#Load your fine-tuned Whisper model and processor
|
22 |
+
model_name = "hackergeek98/tinyyyy_whisper"
|
23 |
+
processor = WhisperProcessor.from_pretrained(model_name)
|
24 |
+
model = WhisperForConditionalGeneration.from_pretrained(model_name)
|
25 |
+
|
26 |
+
#Force the model to transcribe in Persian
|
27 |
+
model.config.forced_decoder_ids = processor.get_decoder_prompt_ids(language="fa", task="transcribe")
|
28 |
+
|
29 |
+
#Move model to GPU if available
|
30 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
31 |
+
model.to(device)
|
32 |
+
import librosa
|
33 |
+
|
34 |
+
def transcribe_audio(audio_file):
|
35 |
+
# Load audio file using librosa (supports multiple formats)
|
36 |
+
audio_data, sampling_rate = librosa.load(audio_file, sr=16000) # Resample to 16kHz
|
37 |
+
|
38 |
+
# Preprocess the audio
|
39 |
+
inputs = processor(audio_data, sampling_rate=sampling_rate, return_tensors="pt").input_features.to(device)
|
40 |
+
|
41 |
+
# Generate transcription
|
42 |
+
with torch.no_grad():
|
43 |
+
predicted_ids = model.generate(inputs)
|
44 |
+
|
45 |
+
# Decode the transcription
|
46 |
+
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
|
47 |
+
return transcription
|
48 |
+
from google.colab import files
|
49 |
+
|
50 |
+
#Upload an audio file
|
51 |
+
uploaded = files.upload()
|
52 |
+
audio_file = list(uploaded.keys())[0]
|
53 |
+
|
54 |
+
#Transcribe the audio
|
55 |
+
transcription = transcribe_audio(audio_file)
|
56 |
+
print("Transcription:", transcription)
|