physics learning
Physics-Inspired-Landscape-Learning-and-Flow-Matching-data
Physics-Inspired Landscape Learning and Flow Matching
LandFlow trajectory supervision
This repository contains the 3,000-trajectory bank used to train the LandFlow
route potential, the fixed CrossDocked split, atom-number metadata, reference
metric cache, and the PAFlow-ready CrossDocked pocket data required by the
reproduction workflow.
The bank contains 50 recorded states per trajectory (150,000 supervised state
examples). It was built from 100 training pockets… See the full description on the dataset page: https://huggingface.co/datasets/NYCU-MLLab/Physics-Inspired-Landscape-Learning-and-Flow-Matching-data.wm-continual-learning-physics
