How Reinforcement Learning and Stable Diffusion Are Being Combined to Simulate Game Worlds
We train all simulation models from a pretrained checkpoint of Stable Diffusion 1.4, unfreezing all U-Net parameters. We use a batch size of 128 ...









