Datasets Audio and Vision Update

Hugging Face has updated the 🤗 Datasets library with new documentation and features for efficiently loading and processing audio and vision (image) datasets.

Key updates are as follows:

• Quickstart update: Includes end-to-end examples for audio and image datasets, guiding data loading and processing methods in both PyTorch and TensorFlow environments.
• Addition of the to_tf_dataset function: Enables easy conversion of datasets into tf.data.Dataset format, allowing them to be used directly for TensorFlow/Keras model training.
• Dedicated guides provided: New dedicated sections tailored to the characteristics of each data modality, offering detailed modality-specific processing methods such as automatic decoding and resampling of audio signals.
• ImageFolder support: Enhanced functionality that allows building image classification datasets instantly using just a folder structure, without writing a separate loading script.