arlbench.core.algorithms.buffers

Replay buffers.

Functions

uniform_sample(state, rng_key, batch_size, ...)

Adapted sample function to support uniform sampling for priorizized buffers.

arlbench.core.algorithms.buffers.uniform_sample(state, rng_key, batch_size, sequence_length, period)[source]

Adapted sample function to support uniform sampling for priorizized buffers.

Parameters:
  • state (PrioritisedTrajectoryBufferState[Experience]) – Buffer state.

  • rng_key (chex.PRNGKey) – Random generator key.

  • batch_size (int) – Sample batch size.

  • sequence_length (int) – Length of trajectory to sample.

  • period (int) – Interval between sampled sequences.

Returns:

Batch of experience.

Return type:

TransitionSample