UmarRamzan/w2v2-bert-urdu

Wav2Vec-Bert-2.0-Urdu

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the Urdu split of the Common Voice 17 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3681
• Wer: 0.2929

Usage Instructions

from transformers import AutoFeatureExtractor, Wav2Vec2BertModel
import torch
from datasets import load_dataset

dataset = load_dataset("hf-internal-testing/librispeech_asr_demo", "clean", split="validation")
dataset = dataset.sort("id")
sampling_rate = dataset.features["audio"].sampling_rate

processor = AutoProcessor.from_pretrained("UmarRamzan/w2v2-bert-urdu")
model = Wav2Vec2BertModel.from_pretrained("UmarRamzan/w2v2-bert-urdu")

# audio file is decoded on the fly
inputs = processor(dataset[0]["audio"]["array"], sampling_rate=sampling_rate, return_tensors="pt")
with torch.no_grad():
    outputs = model(**inputs)

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-06
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4362	0.1695	50	0.4144	0.3213
0.3776	0.3390	100	0.4029	0.3137
0.3918	0.5085	150	0.4095	0.3060
0.3968	0.6780	200	0.3961	0.3060
0.3685	0.8475	250	0.3681	0.2929

Framework versions

• Transformers 4.40.2
• Pytorch 2.2.1+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1