wav2vec2-large-xlsr-moroccan-darija

Wav2Vec2-Large-XLSR-53-Moroccan-Darija

wav2vec2-large-xlsr-53 new model

• Fine-tuned on 57 hours of labeled Darija Audios extracted from MDVC (https://ijeecs.iaescore.com/index.php/IJEECS/article/view/35709) which contains more than 1000 hours of Moroccan Darija "ary".
• Fine-tuning is ongoing 24/7 to enhance accuracy.
• We are consistently adding data to the model every day (We prefer not to add all MDVC Corpus at once as we are trying to standardize more and more the way we write the Moroccan Darija).

<table><thead><tr><th><strong>Training Loss</strong></th> <th><strong>Validation</strong></th> <th><strong>Loss Wer</strong></th></tr></thead> <tbody><tr> <td>0.121300</td> <td>0.103430</td> <td>0.084904</td> </tr> </tbody></table>

Usage

The model can be used directly as follows:

import librosa
import torch
from transformers import Wav2Vec2CTCTokenizer, Wav2Vec2ForCTC, Wav2Vec2Processor, TrainingArguments, Wav2Vec2FeatureExtractor, Trainer

tokenizer = Wav2Vec2CTCTokenizer("./vocab.json", unk_token="[UNK]", pad_token="[PAD]", word_delimiter_token="|")
processor = Wav2Vec2Processor.from_pretrained('boumehdi/wav2vec2-large-xlsr-moroccan-darija', tokenizer=tokenizer)
model=Wav2Vec2ForCTC.from_pretrained('boumehdi/wav2vec2-large-xlsr-moroccan-darija')


# load the audio data (use your own wav file here!)
input_audio, sr = librosa.load('file.wav', sr=16000)

# tokenize
input_values = processor(input_audio, return_tensors="pt", padding=True).input_values

# retrieve logits
logits = model(input_values).logits

tokens = torch.argmax(logits, axis=-1)

# decode using n-gram
transcription = tokenizer.batch_decode(tokens)

# print the output
print(transcription)

Output: قالت ليا هاد السيد هادا ما كاينش بحالو

email: [email protected]

BOUMEHDI Ahmed