torchrl.trainers.algorithms.configs.data.PrioritizedSamplerConfig¶
- class torchrl.trainers.algorithms.configs.data.PrioritizedSamplerConfig(_target_: str = 'torchrl.data.replay_buffers.PrioritizedSampler', max_capacity: int | None = None, alpha: float | None = None, beta: float | None = None, eps: float | None = None, reduction: str | None = None)[source]¶
Configuration for prioritized sampling from replay buffer.