Completed
My crazy idea of Metatoken and ICL NVIDA
Class Central Classrooms beta
YouTube videos curated by Class Central.
Classroom Contents
NVIDIA HYMBA: A Hybrid-Head Architecture for Small Language Models with MetaTokens
Automatically move to the next video in the Classroom when playback concludes
- 1 New NVIDIA HYMBA LLM
- 2 Inference run w Test time training
- 3 Transformer in parallel w MAMBA
- 4 Metatoken introduced
- 5 Task specific Metatoken
- 6 MetaTokens explained in detail
- 7 NVIDIA Hymba beats Llama 3.2 3B
- 8 Attention map Entropy per Head
- 9 Key Value Cache in Transformer & Mamba
- 10 My crazy idea of Metatoken and ICL NVIDA
