H Net Dynamic Chunking Hierarchical Sequence

Eliminate fixed tokenization by learning data-dependent segmentation jointly with the model through dynamic chunking, matching BPE-tokenized Transformers at equivalent compute while showing improved robustness and better downstream task performance without vocabulary constraints.

adu2021 Updated 7 repo stars

File contents

adu2021/skillxiv commit b48a1906b4

Frequently asked questions

npx get-skill add adu2021/h-net-dynamic-chunking-hierarchical-sequence