algorithms.dataset package

Submodules

algorithms.dataset.CriticDataset module

class algorithms.dataset.CriticDataset.CriticDataset(data, shuffle=True, **kwargs)[source]

Bases: algorithms.dataset.CriticDataset.Dataset

batches(samples_data, store=True)[source]
Parameters

  • samples_data – batches of sampled observation action pairs

  • store – indicator of whether store sampled data in replay buffer or not

Returns

yielded batch

class algorithms.dataset.CriticDataset.Dataset(data, batch_size, action_normalizer=None, observation_normalizer=None, replay_memory=None, recurrent=False, flat_recurrent=False, use_random_scaling=False, random_scale_factor=0.2, use_random_noise=False, random_noise_factor=0.003)[source]

Bases: object

batches(samples_data, store=True)[source]
algorithms.dataset.CriticDataset.select_batch_idxs(start_idx, batch_size, min_idx, max_idx)[source]
Parameters

  • start_idx – starting index

  • batch_size – batch size

  • min_idx – minimum index

  • max_idx – maximum index

Returns

the selected list of indexes

algorithms.dataset.utils module

class algorithms.dataset.utils.KeyValueReplayMemory(maxsize=None)[source]

Bases: object

add(keys, values)[source]

Adds keys from values to memory Args:

  • keys: the keys to add, list of hashable

  • values: dict containing each key in keys

sample(keys, size)[source]

Sample a batch of size for each key and return as a dict Args:

  • keys: list of keys

  • size: number of samples to select

algorithms.dataset.utils.compute_n_batches(n_samples, batch_size)[source]
Parameters

  • n_samples – how many samples

  • batch_size – how many samples in one batch

Returns

how many batches we need

algorithms.dataset.utils.load_dataset(filepath, maxsize=None)[source]
Parameters

  • filepath – file path of the data set

  • maxsize – max size default is None

Returns

loaded data set

algorithms.dataset.utils.pad_tensor(x, max_len, axis)[source]
Parameters

  • x – given x

  • max_len – max length

  • axis – the axis on which to be padded

Returns

padded x

Module contents