Metadata-Version: 2.1
Name: rlkit
Version: 0.2.1.dev0
License: MIT License
License-File: LICENSE

# Reinforcement-Learning-as-One-Class-Classification

This is the official implementation of Reinforcement-Learning-as-One-Class-Classification(RL-OCC) that accompanies the research paper, Reinforcement-Learning-as-One-Class-Classification (will be submitted).

Reinforcement-Learning-as-One-Class-Classification, blah blah
## Getting Started
This codebase is built off of the official implementation of BEAR (https://github.com/rail-berkeley/d4rl_evaluations/tree/master/bear) and rlkit (https://github.com/vitchyr/rlkit/). In order to run UWAC, follow the installation instructions for rlkit as shown below, then install D4RL(https://github.com/rail-berkeley/d4rl).


############고치자###########
Then in order to run UWAC, an example command is:
```
python examples/UWAC_hdf5_d4rl.py --env='halfcheetah-medium-v0' --policy_lr=1e-4 --num_samples=100
```
where, `env` refers to a d4rl environment, `policy_lr` is the policy learning rate.

To run BEAR (baseline), an example command is:
```
python examples/bear_hdf5_d4rl.py --env='halfcheetah-medium-v0' --policy_lr=1e-4 --num_samples=100
```

See readme from RLkit (https://github.com/vitchyr/rlkit) for troubleshooting with respect to RLkit problems.

# License
This code is released under the [LICENSE](LICENSE) terms.
