ray.rllib.policy.sample_batch.SampleBatch#

class ray.rllib.policy.sample_batch.SampleBatch(*args, **kwargs)[source]#

Bases: dict

Wrapper around a dictionary with string keys and array-like values.

For example, {“obs”: [1, 2, 3], “reward”: [0, -1, 1]} is a batch of three samples, each with an “obs” and “reward” attribute.

Methods

`__init__`	Constructs a sample batch (same params as dict constructor).
`agent_steps`	Returns the same as len(self) (number of steps in this batch).
`as_multi_agent`	Returns the respective MultiAgentBatch using DEFAULT_POLICY_ID.
`clear`
`columns`	Returns a list of the batch-data in the specified columns.
`compress`	Compresses the data buffers (by column) in place.
`concat`	Concatenates `other` to this one and returns a new SampleBatch.
`copy`	Creates a deep or shallow copy of this SampleBatch and returns it.
`decompress_if_needed`	Decompresses data buffers (per column if not compressed) in place.
`env_steps`	Returns the same as len(self) (number of steps in this batch).
`fromkeys`	Create a new dictionary with keys from iterable and values set to value.
`get`	Returns one column (by key) from the data or a default value.
`get_single_step_input_dict`	Creates single ts SampleBatch at given index from `self`.
`is_single_trajectory`	Returns True if this SampleBatch only contains one trajectory.
`is_terminated_or_truncated`	Returns True if `self` is either terminated or truncated at idx -1.
`items`
`keys`
`pop`	If the key is not found, return the default if given; otherwise, raise a KeyError.
`popitem`	Remove and return a (key, value) pair as a 2-tuple.
`right_zero_pad`	Right (adding zeros at end) zero-pads this SampleBatch in-place.
`rows`	Returns an iterator over data rows, i.e. dicts with column values.
`set_get_interceptor`	Sets a function to be called on every getitem.
`set_training`	Sets the `is_training` flag for this SampleBatch.
`setdefault`	Insert key with a value of default if key is not in the dictionary.
`shuffle`	Shuffles the rows of this batch in-place.
`size_bytes`	Returns sum over number of bytes of all data buffers.
`slice`	Returns a slice of the row data of this batch (w/o copying).
`split_by_episode`	Splits by `eps_id` column and returns list of new batches.
`timeslices`	Returns SampleBatches, each one representing a k-slice of this one.
`to_device`	TODO: transfer batch to given device as framework tensor.
`update`	If E is present and has a .keys() method, then does: for k in E: D[k] = E[k] If E is present and lacks a .keys() method, then does: for k, v in E: D[k] = v In either case, this is followed by: for k in F: D[k] = F[k]
`values`

Attributes

`ACTIONS`
`ACTION_DIST`
`ACTION_DIST_INPUTS`
`ACTION_LOGP`
`ACTION_PROB`
`AGENT_INDEX`
`ATTENTION_MASKS`
`CUR_OBS`
`DONES`
`ENV_ID`
`EPS_ID`
`INFOS`
`NEXT_OBS`
`OBS`
`OBS_EMBEDS`
`PREV_ACTIONS`
`PREV_REWARDS`
`RETURNS_TO_GO`
`REWARDS`
`SEQ_LENS`
`T`
`TERMINATEDS`
`TRUNCATEDS`
`UNROLL_ID`
`VALUES_BOOTSTRAPPED`
`VF_PREDS`
`is_training`