Easy-to-use D4RL offline dataset loader
Project description
just-d4rl
Easy-to-use D4RL offline dataset loader, focused solely on downloading and providing D4RL datasets without dependencies on gym or gymnasium.
Key Features
- Downloads and provides D4RL offline datasets
- No dependencies on gym or gymnasium
- Lightweight and focused functionality
- Easy integration with PyTorch
Installation
Install from PyPI:
pip install just-d4rl
Usage
from just_d4rl import D4RLDataset, d4rl_offline_dataset, d4rl_score_normalizer
# Download and load a D4RL dataset
d4rl_dataset = d4rl_offline_dataset("hopper-medium-v2")
# Example datasets
d4rl_dataset = d4rl_offline_dataset("walker2d-random-v2")
d4rl_dataset = d4rl_offline_dataset("halfcheetah-medium-expert-v2")
d4rl_dataset = d4rl_offline_dataset("antmaze-umaze-v2")
dataset = d4rl_dataset
dataset['observations'].shape, dataset['actions'].shape, dataset['rewards'].shape, dataset['next_observations'].shape, dataset['terminals'].shape
# ((1000000, 11), (1000000, 3), (1000000,), (1000000, 11), (1000000,))
# Create a PyTorch Dataset
d4rl_dataset = d4rl_offline_dataset("hopper-medium-v2")
dataset = D4RLDataset(d4rl_dataset)
# Get a batch of data
batch = dataset[-16:]
batch["observation"].shape, batch["action"].shape, batch["reward"].shape, batch["next_observation"].shape, batch["terminal"].shape
# (torch.Size([16, 11]), torch.Size([16, 3]), torch.Size([16]), torch.Size([16, 11]), torch.Size([16]))
returns = np.random.rand(16, 1)
get_normalized_score = d4rl_score_normalizer("hopper-medium-v2")
get_normalized_score(returns)
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
just_d4rl-0.2407.3.tar.gz
(9.2 kB
view hashes)
Built Distribution
Close
Hashes for just_d4rl-0.2407.3-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 83be6b9abe711c1109396dc14184315f085a17519f5c7dcabcddd7fdfe69a024 |
|
MD5 | 78f18e8c561513ff4762bf9d4b8b538e |
|
BLAKE2b-256 | 0edae3fdba9f48e1128fc5f5335baedec20d0216db54870280f165cedc856e64 |