metadata
license: apache-2.0
Dataset Card for Pong-v4-expert-MCTS
Table of Contents
Dataset Description
This dataset includes 8 episodes of pong-v4 environment. The expert policy is EfficientZero, which is able to generate MCTS hidden states.
Supported Tasks and Leaderboards
Dataset Structure
Data Instances
A data point comprises tuples of sequences of (observations, actions, hidden_states):
{
"obs":datasets.Array3D(),
"actions":int,
"hidden_state":datasets.Array3D(),
}
Data Fields
obs
: An Array3D containing observations from 8 trajectories of an evaluated agent. The data type is uint8 and each value is in 0 to 255.actions
: An integer containing actions from 8 trajectories of an evaluated agent. This value is from 0 to 5.hidden_state
: An Array3D containing corresponding hidden states generated by EfficientZero, from 8 trajectories of an evaluated agent. The data type is float32.
Data Splits
There is only a training set for this dataset, as evaluation is undertaken by interacting with a simulator.
Dataset Creation
Curation Rationale
- TBD
Source Data
Initial Data Collection and Normalization
- TBD
Who are the source language producers?
- TBD
Annotations
- TBD
Considerations for Using the Data
Known Limitations
- TBD
Additional Information
Licensing Information
- TBD
Citation Information
TBD
Contributions
Thanks to @test, for adding this dataset.