LiaAira/Taylor-Data
π Taylor-Data: Large-Scale Neural Dynamics & Optimization Dataset (V4) Taylor-Data is a comprehensive, large-scale dataset containing 2 Billion parameter transitions and loss landscape interactions collected across diverse neural network architectures, optimization regimes, and multi-scale horizons. It is designed for training World Models of Neural Dynamics, Meta-Optimizers, Quasi-Newton Credit Assignment Networks, and Continual Learning Controllers. π Overview &β¦ See the full description on the dataset page: https://huggingface.co/datasets/LiaAira/Taylor-Data.
Conversations for this repository live on Hugging Face.
CoolFace shows imported repositories read-only. Posting into someone elseβs repository from here would need an authorised integration and the account holderβs consent, so the link goes to the source instead.
Open discussions on Hugging Face