Data Augmentation

[twinkle485]

Participants can increase the diversity of their training data (e.g., vehicles, animals, facial images, etc.) by applying transformations like cropping, flipping, or noise injection. These augmentations help make the model more robust by exposing it to variations, allowing LoRA layers to learn better representations, even with a limited dataset.
Please ensure you have read the guidelines in CONTRIBUTING.md and CODE_OF_CONDUCT.md before proceeding.