Modelpublic

megaaziib/wav2vec2-large-xlsr-indonesian-safetensors

sourceHugging Faceapache-2.0updated 6mo agoView on Hugging Face

1likes33downloads

Model Card

wav2vec2-large-xlsr-indonesian (Safetensor Variant)

This model is a conversion of the original cahya/wav2vec2-large-xlsr-indonesian into the Safetensors format. Safetensors is a specialized format for storing tensors that is secure, fast, and facilitates efficient loading (lazy loading).

Model Details

Model Description

—Developed by: Cahya Wirawan (Original Model)
—Converted & Shared by: Megaaziib
—Model type: XLSR-Wav2Vec2
—Language(s) (NLP): Indonesian (id)
—License: Apache 2.0
—Finetuned from model: facebook/wav2vec2-large-xlsr-53

Model Sources

—Original Repository: cahya/wav2vec2-large-xlsr-indonesian

Uses

Direct Use

This model is intended for Automatic Speech Recognition (ASR) for the Indonesian language. It can be used directly for transcribing Indonesian audio files or as a backbone for further fine-tuning on specific Indonesian dialects or domains (e.g., medical, legal).

Out-of-Scope Use

The model may not perform well on:

—Low-quality audio or heavy background noise.
—Non-Indonesian languages.
—Extremely thick regional accents not represented in the original Common Voice or training datasets.

Bias, Risks, and Limitations

This model inherits the limitations of the original XLSR-53 architecture and the specific Indonesian training data used by Cahya. Users should be aware of potential biases toward formal Indonesian (Bahasa Baku) versus informal slang.

How to Get Started with the Model

You can use this model directly with the transformers library. Ensure you have safetensors installed.

python

from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor

model_id = "megaaziib/wav2vec2-large-xlsr-indonesian" # Replace with your actual repo name if different

processor = Wav2Vec2Processor.from_pretrained(model_id)
model = Wav2Vec2ForCTC.from_pretrained(model_id, use_safetensors=True)