tutorials.examples.train_hypergrid_buffer¶
Example script for training a GFlowNet using replay buffers.
This script demonstrates two approaches for off-policy training: 1. Trajectory buffer: Stores and samples entire trajectories for training. 2. Terminating state buffer: Stores terminating states, from which backward trajectories are sampled.
Both buffer types can be selected to improve training efficiency and diversity.
Attributes¶
Functions¶
|
Module Contents¶
- tutorials.examples.train_hypergrid_buffer.main(args)¶
- tutorials.examples.train_hypergrid_buffer.parser¶