Closed aaprasad closed 9 months ago
Hi, really helpful implementation! I know RoPE was originally designed for NLP tasks but I was wondering how it might be extended to other domains like computer vision with images?
there's an example in the readme for axial rotary embeddings, allowing for images and video support
Hi, really helpful implementation! I know RoPE was originally designed for NLP tasks but I was wondering how it might be extended to other domains like computer vision with images?