LLM “Patchnization”
LLM “Patchnization”
Minimal, clean code for video/image "patchnization" - a process commonly applied in tokenizing visual data for use in a Transformer encoder.
Code: https://github.com/Jaykef/min-patchnizer
The code above, first extracts still images (frames) from a video, splits the image frames into smaller fixed-size patches, linearly embeds each of them, adds position embeddings and then saves the resulting sequence of vectors for use in a Vision Transformer encoder. I tried training the resulting sequence vectors with Karpathy's minbpe and it took ~30s per frame to tokenize. The whole "patchnization" took ~10s for a 20s video on my M2 Air.
The files in the repo work as follows:
1. patchnizer.py: Holds code for simple implemenatation of the three stages involved (extract_image_frames from video, reduce image_frames_to_patches of fixed sizes 16x16 pixels, then linearly_embed_patches into a 1D vector sequence with additional position embeddings.
2. patchnize.py: performs the whole process with custom configs (patch_size, created dirs, video - I am using the "dogs playing in snow" video by sora).
3. train.py: Trains the resulting one-dimensional vector sequence (linear_patch_embeddings + position_embeddings) on Karpathy's minbpe (a minimal implementation of the byte-pair encoding algorithm).
4. check.py: Checks to see if the patch embeddings match the original image patches - this basically just do the reverse of linear embedding.
The whole process builds on the approach introduced in the Vision Transformer paper: "An image is worth 16x16 words: Transformers for image recognition at scale."
Paper: https://arxiv.org/abs/2010.11929
#llm #sora #nlp #openai