Skip to content

Latest commit

 

History

History
139 lines (116 loc) · 5.35 KB

README.md

File metadata and controls

139 lines (116 loc) · 5.35 KB
Raw

Source code of the paper "Beyond Optimism: Exploration With Partially Observable Rewards" (Arxiv, NeurIPS).

Install and Examples

To install and use our environments, run

pip install -r requirements.txt
cd src/gym_gridworlds
pip install -e .

Run python and then

import gymnasium
import gym_gridworlds
env = gymnasium.make("Gym-Gridworlds/Penalty-3x3-v0", render_mode="human")
env.reset()
env.step(1) # DOWN
env.step(4) # STAY
env.render()

to render the Penalty-3x3-v0 (left figure), and

import gymnasium
import gym_gridworlds
env = gymnasium.make("Gym-Gridworlds/Full-5x5-v0", render_mode="human")
env.reset()
env.step(1) # DOWN
env.render()

to render the Full-5x5-v0 (right figure).

Gridworld Penalty       Gridworld Full

  • Black tiles are empty,
  • Black tiles with gray arrows are tiles where the agent can move only in one direction (other actions will fail),
  • Red tiles give negative rewards,
  • Green tiles give positive rewards (the brighter, the higher),
  • Yellow tiles are quicksands, where all actions will fail with 90% probability,
  • The agent is the blue circle,
  • The orange arrow denotes the agent's last action,
  • The orange dot denotes that the agent did not try to move with its last action.

It is also possible to add noise to the transition and the reward functions. For example, the following environment

import gymnasium
import gym_gridworlds
env = gymnasium.make("Gym-Gridworlds/Full-5x5-v0", random_action_prob=0.1, reward_noise_std=0.05)
  • Performs a random action with 10% probability (regardless of what the agent wants to do),
  • Adds Gaussian noise with 0.05 standard deviation to the reward.

Hydra Configs

We use Hydra to configure our experiments.
Hyperparameters and other settings are defined in YAML files in the configs/ folder.
Most of the configuration is self-explanatory. Some keys you may need to change are the following:

  • WandB settings and Hydra log directories in configs/default.yaml,
  • Folder experiment.datadir in configs/default.yaml (where npy data is saved),
  • Folder experiment.debugdir in configs/default.yaml (where agent pics are saved),
  • Learning rate, epsilon decay, Q-function initial values, and other agent parameters are configs/agent/,
  • Training/testing setup is in configs/environment/.

Quick Run

python main.py experiment.debugdir=debug environment=penalty monitor=full

This will save pics to easily debug a run.
Everything will be saved in debug/, in subfolders depending on the Git commit and environment name.
For example, you will find these two heatmaps, representing the state-action visitation count and the Q-function.

  • States are on the y-axis,
  • Actions are on the x-axis,
  • Cells outlined in red are for the actions with the highest value in each state.

Visit Count       Q-Function

Run the following script to see the optimal Q-function

python fqi.py experiment.debugdir=debug environment=penalty monitor=full

Sweeps

For a sweep over multiple jobs in parallel with Joblib, run

python main.py -m hydra/launcher=joblib hydra/sweeper=test

Custom sweeps are defined in configs/hydra/sweeper/.
You can further customize a sweep via command line. For example,

python main.py -m hydra/launcher=joblib hydra/sweeper=test experiment.rng_seed="range(0, 10)" hydra.launcher.verbose=1000

Configs in configs/hydra/sweeper/ hide the training progress bar of the agent, so we suggest to pass hydra.launcher.verbose=1000 to show the progress of the sweep.

If you have access to a SLURM-based cluster, you can submit multiple jobs, each running a chunk of the sweep with Joblib. Refer to submitit_jobs.py for an example.

Plot Data From Sweeps

Experiments will save the expected discounted return of the ε-greedy (training) and greedy (testing) policies, along with other stats, in a npz file (default dir is data/, followed by the hash of the Git commit).
If you want to zip the whole data folder, run

find data -type f -name "*.npz" -print0 | tar -czvf data.tar.gz --null -T -

To plot expected return curves, use plot_results.py. This script takes two arguments:

  • -c is the config file that defines where to save plots, axes limits, axes ticks, what algorithms to show, and so on. Default configs are located in configs/plots/.
  • -f is the folder where data from the sweep is located.
    A separate script plot_legend.py saves the legend in a separate pic.

For example, running

python plot_results.py -c configs/plots/test.py -f data_test/GIT_HASH/
python plot_legend.py

Will generate many plots like these, and save them in the same folder passed with -f.

Ex. 1       Ex. 2