train.replay_buffer
train.replay_buffer
Circular replay buffer for training positions.
Classes
| Name | Description |
|---|---|
| ReplayBuffer | Stores training positions and samples random batches. |
ReplayBuffer
train.replay_buffer.ReplayBuffer(max_size=100000)Stores training positions and samples random batches.
Methods
| Name | Description |
|---|---|
| sample_batch | Sample a random batch as PyTorch tensors. |
sample_batch
train.replay_buffer.ReplayBuffer.sample_batch(batch_size)Sample a random batch as PyTorch tensors.
Returns: obs_batch: (batch, 16, 2, 12) policy_batch: (batch, 1352) value_batch: (batch,)