README.md

---
language:
- en
license: apache-2.0
size_categories:
- 10M<n<100M
task_categories:
- reinforcement-learning
pretty_name: Procgen Benchmark Dataset
configs:
- config_name: bigfish
  data_files:
  - split: train
    path: bigfish/train-*
  - split: test
    path: bigfish/test-*
- config_name: bossfight
  data_files:
  - split: train
    path: bossfight/train-*
  - split: test
    path: bossfight/test-*
- config_name: caveflyer
  data_files:
  - split: train
    path: caveflyer/train-*
  - split: test
    path: caveflyer/test-*
- config_name: chaser
  data_files:
  - split: train
    path: chaser/train-*
  - split: test
    path: chaser/test-*
- config_name: climber
  data_files:
  - split: train
    path: climber/train-*
  - split: test
    path: climber/test-*
- config_name: coinrun
  data_files:
  - split: train
    path: coinrun/train-*
  - split: test
    path: coinrun/test-*
- config_name: dodgeball
  data_files:
  - split: train
    path: dodgeball/train-*
  - split: test
    path: dodgeball/test-*
- config_name: fruitbot
  data_files:
  - split: train
    path: fruitbot/train-*
  - split: test
    path: fruitbot/test-*
- config_name: heist
  data_files:
  - split: train
    path: heist/train-*
  - split: test
    path: heist/test-*
- config_name: jumper
  data_files:
  - split: train
    path: jumper/train-*
  - split: test
    path: jumper/test-*
- config_name: leaper
  data_files:
  - split: train
    path: leaper/train-*
  - split: test
    path: leaper/test-*
- config_name: maze
  data_files:
  - split: train
    path: maze/train-*
  - split: test
    path: maze/test-*
- config_name: miner
  data_files:
  - split: train
    path: miner/train-*
  - split: test
    path: miner/test-*
- config_name: ninja
  data_files:
  - split: train
    path: ninja/train-*
  - split: test
    path: ninja/test-*
- config_name: plunder
  data_files:
  - split: train
    path: plunder/train-*
  - split: test
    path: plunder/test-*
- config_name: starpilot
  data_files:
  - split: train
    path: starpilot/train-*
  - split: test
    path: starpilot/test-*
tags:
- procgen
- bigfish
- benchmark
- openai
- bossfight
- caveflyer
- chaser
- climber
- dodgeball
- fruitbot
- heist
- jumper
- leaper
- maze
- miner
- ninja
- plunder
- starpilot
dataset_info:
- config_name: bigfish
  features:
  - name: observation
    dtype:
      image:
        mode: RGB
  - name: action
    dtype: int32
  - name: reward
    dtype: float32
  - name: terminated
    dtype: bool
  - name: truncated
    dtype: bool
  splits:
  - name: train
    num_bytes: 52167166912.0
    num_examples: 9000000
  - name: test
    num_bytes: 5791205186.0
    num_examples: 1000000
  download_size: 57881994479
  dataset_size: 57958372098.0
- config_name: bossfight
  features:
  - name: observation
    dtype:
      image:
        mode: RGB
  - name: action
    dtype: int32
  - name: reward
    dtype: float32
  - name: terminated
    dtype: bool
  - name: truncated
    dtype: bool
  splits:
  - name: train
    num_bytes: 66990875098.0
    num_examples: 9000000
  - name: test
    num_bytes: 7422341538.0
    num_examples: 1000000
  download_size: 74364918274
  dataset_size: 74413216636.0
- config_name: caveflyer
  features:
  - name: observation
    dtype:
      image:
        mode: RGB
  - name: action
    dtype: int32
  - name: reward
    dtype: float32
  - name: terminated
    dtype: bool
  - name: truncated
    dtype: bool
  splits:
  - name: train
    num_bytes: 43639897909.0
    num_examples: 9000000
  - name: test
    num_bytes: 4873680436.0
    num_examples: 1000000
  download_size: 48700939902
  dataset_size: 48513578345.0
- config_name: chaser
  features:
  - name: observation
    dtype:
      image:
        mode: RGB
  - name: action
    dtype: int32
  - name: reward
    dtype: float32
  - name: terminated
    dtype: bool
  - name: truncated
    dtype: bool
  splits:
  - name: train
    num_bytes: 51447214089.0
    num_examples: 9000000
  - name: test
    num_bytes: 5714598297.0
    num_examples: 1000000
  download_size: 55552919888
  dataset_size: 57161812386.0
- config_name: climber
  features:
  - name: observation
    dtype:
      image:
        mode: RGB
  - name: action
    dtype: int32
  - name: reward
    dtype: float32
  - name: terminated
    dtype: bool
  - name: truncated
    dtype: bool
  splits:
  - name: train
    num_bytes: 24375527245.0
    num_examples: 9000000
  - name: test
    num_bytes: 2690264702.0
    num_examples: 1000000
  download_size: 26574301930
  dataset_size: 27065791947.0
- config_name: coinrun
  features:
  - name: observation
    dtype:
      image:
        mode: RGB
  - name: action
    dtype: int32
  - name: reward
    dtype: float32
  - name: terminated
    dtype: bool
  - name: truncated
    dtype: bool
  splits:
  - name: train
    num_bytes: 16788785439.0
    num_examples: 9000000
  - name: test
    num_bytes: 1875181202.0
    num_examples: 1000000
  download_size: 18088978913
  dataset_size: 18663966641.0
- config_name: dodgeball
  features:
  - name: observation
    dtype:
      image:
        mode: RGB
  - name: action
    dtype: int32
  - name: reward
    dtype: float32
  - name: terminated
    dtype: bool
  - name: truncated
    dtype: bool
  splits:
  - name: train
    num_bytes: 62421409783.0
    num_examples: 9000000
  - name: test
    num_bytes: 6939014458.0
    num_examples: 1000000
  download_size: 69244806537
  dataset_size: 69360424241.0
- config_name: fruitbot
  features:
  - name: observation
    dtype:
      image:
        mode: RGB
  - name: action
    dtype: int32
  - name: reward
    dtype: float32
  - name: terminated
    dtype: bool
  - name: truncated
    dtype: bool
  splits:
  - name: train
    num_bytes: 59648385662.0
    num_examples: 9000000
  - name: test
    num_bytes: 6610963384.0
    num_examples: 1000000
  download_size: 66466587233
  dataset_size: 66259349046.0
- config_name: heist
  features:
  - name: observation
    dtype:
      image:
        mode: RGB
  - name: action
    dtype: int32
  - name: reward
    dtype: float32
  - name: terminated
    dtype: bool
  - name: truncated
    dtype: bool
  splits:
  - name: train
    num_bytes: 55385943063.0
    num_examples: 9000000
  download_size: 45218795172
  dataset_size: 55385943063.0
- config_name: jumper
  features:
  - name: observation
    dtype:
      image:
        mode: RGB
  - name: action
    dtype: int32
  - name: reward
    dtype: float32
  - name: terminated
    dtype: bool
  - name: truncated
    dtype: bool
  splits:
  - name: train
    num_bytes: 23713574993.0
    num_examples: 9000000
  - name: test
    num_bytes: 2596720354.0
    num_examples: 1000000
  download_size: 26160892415
  dataset_size: 26310295347.0
- config_name: leaper
  features:
  - name: observation
    dtype:
      image:
        mode: RGB
  - name: action
    dtype: int32
  - name: reward
    dtype: float32
  - name: terminated
    dtype: bool
  - name: truncated
    dtype: bool
  splits:
  - name: train
    num_bytes: 44825702164.0
    num_examples: 9000000
  - name: test
    num_bytes: 4970059656.0
    num_examples: 1000000
  download_size: 48913285349
  dataset_size: 49795761820.0
- config_name: maze
  features:
  - name: observation
    dtype:
      image:
        mode: RGB
  - name: action
    dtype: int32
  - name: reward
    dtype: float32
  - name: terminated
    dtype: bool
  - name: truncated
    dtype: bool
  splits:
  - name: train
    num_bytes: 49276863929.0
    num_examples: 9000000
  - name: test
    num_bytes: 5482794667.0
    num_examples: 1000000
  download_size: 38922389228
  dataset_size: 54759658596.0
- config_name: miner
  features:
  - name: observation
    dtype:
      image:
        mode: RGB
  - name: action
    dtype: int32
  - name: reward
    dtype: float32
  - name: terminated
    dtype: bool
  - name: truncated
    dtype: bool
  splits:
  - name: train
    num_bytes: 44058761019.0
    num_examples: 9000000
  - name: test
    num_bytes: 4876567632.0
    num_examples: 1000000
  download_size: 17739617811
  dataset_size: 48935328651.0
- config_name: ninja
  features:
  - name: observation
    dtype:
      image:
        mode: RGB
  - name: action
    dtype: int32
  - name: reward
    dtype: float32
  - name: terminated
    dtype: bool
  - name: truncated
    dtype: bool
  splits:
  - name: train
    num_bytes: 28875440305.0
    num_examples: 9000000
  - name: test
    num_bytes: 3207526456.0
    num_examples: 1000000
  download_size: 32127321141
  dataset_size: 32082966761.0
- config_name: plunder
  features:
  - name: observation
    dtype:
      image:
        mode: RGB
  - name: action
    dtype: int32
  - name: reward
    dtype: float32
  - name: terminated
    dtype: bool
  - name: truncated
    dtype: bool
  splits:
  - name: train
    num_bytes: 58061310067.0
    num_examples: 9000000
  - name: test
    num_bytes: 6454993435.0
    num_examples: 1000000
  download_size: 64177454509
  dataset_size: 64516303502.0
- config_name: starpilot
  features:
  - name: observation
    dtype:
      image:
        mode: RGB
  - name: action
    dtype: int32
  - name: reward
    dtype: float32
  - name: terminated
    dtype: bool
  - name: truncated
    dtype: bool
  splits:
  - name: train
    num_bytes: 56365605339.0
    num_examples: 9000000
  - name: test
    num_bytes: 6192523494.0
    num_examples: 1000000
  download_size: 62629456972
  dataset_size: 62558128833.0
---
# Procgen Benchmark
This dataset contains expert trajectories generated by a [PPO](https://arxiv.org/abs/1707.06347) reinforcement learning agent trained on each of the 16 procedurally-generated gym environments from the [Procgen Benchmark](https://openai.com/index/procgen-benchmark/). The environments were created on `distribution_mode=easy` and with unlimited levels.

Disclaimer: This is not an official repository from OpenAI.

## Dataset Usage

Regular usage (for environment bigfish):
```python
from datasets import load_dataset
train_dataset = load_dataset("EpicPinkPenguin/procgen", name="bigfish", split="train")
test_dataset = load_dataset("EpicPinkPenguin/procgen", name="bigfish", split="test")
```

Usage with PyTorch (for environment bossfight):
```python
from datasets import load_dataset
train_dataset = load_dataset("EpicPinkPenguin/procgen", name="bossfight", split="train").with_format("torch")
test_dataset = load_dataset("EpicPinkPenguin/procgen", name="bossfight", split="test").with_format("torch")
```

## Agent Performance
The PPO RL agent was trained for 25M steps on each environment and obtained the following final performance metrics on the evaluation environment. These values are attain or surpass the performance described in "Easy Difficulty Baseline Results" in Appendix I of the paper.

| Environment |  Steps (Train)  |  Steps (Test)  | Return | Observation |
|:------------|:----------------|:---------------|:-------|:------------|
| bigfish     | 9,000,000       | 1,000,000      | 33.79  | <video controls autoplay loop src="https://cdn-uploads.huggingface.co/production/uploads/633c1daf31c06121a58f2df9/lHQXBqLdoWicXlt68I9QX.mp4"></video>  |
| bossfight   | 9,000,000       | 1,000,000      | 11.47  | <video controls autoplay loop src="https://cdn-uploads.huggingface.co/production/uploads/633c1daf31c06121a58f2df9/LPoafGi4YBWqqkuFlEN_l.mp4"></video>  |
| caveflyer   | 9,000,000       | 1,000,000      | 09.42  | <video controls autoplay loop src="https://cdn-uploads.huggingface.co/production/uploads/633c1daf31c06121a58f2df9/XVqRwu_9yfX4ECQc4At4G.mp4"></video>  |
| chaser      | 9,000,000       | 1,000,000      | 10.55  | <video controls autoplay loop src="https://cdn-uploads.huggingface.co/production/uploads/633c1daf31c06121a58f2df9/FIKVv48SThqiC1Z2PYQ7U.mp4"></video>  |
| climber     | 9,000,000       | 1,000,000      | 11.30  | <video controls autoplay loop src="https://cdn-uploads.huggingface.co/production/uploads/633c1daf31c06121a58f2df9/XJQlA7IyF9_gwUiw-FkND.mp4"></video>  |
| coinrun     | 9,000,000       | 1,000,000      | 09.02  | <video controls autoplay loop src="https://cdn-uploads.huggingface.co/production/uploads/633c1daf31c06121a58f2df9/Ucv3HZttewMRQzTL8r_Tw.mp4"></video>  |
| dodgeball   | 9,000,000       | 1,000,000      | 13.90  | <video controls autoplay loop src="https://cdn-uploads.huggingface.co/production/uploads/633c1daf31c06121a58f2df9/5HetbKuXBpO-v1jcVyLTU.mp4"></video>  |
| fruitbot    | 9,000,000       | 1,000,000      | 31.58  | <video controls autoplay loop src="https://cdn-uploads.huggingface.co/production/uploads/633c1daf31c06121a58f2df9/zKCyxXvauXjUac-5kEAWz.mp4"></video>  |
| heist       | 9,000,000       | 1,000,000      | 08.32  | <video controls autoplay loop src="https://cdn-uploads.huggingface.co/production/uploads/633c1daf31c06121a58f2df9/AdZ6XNmUN5_00BKd9BN8R.mp4"></video>  |
| jumper      | 9,000,000       | 1,000,000      | 08.10  | <video controls autoplay loop src="https://cdn-uploads.huggingface.co/production/uploads/633c1daf31c06121a58f2df9/s5k31gWK2Vc6Lp6QVzQXA.mp4"></video>  |
| leaper      | 9,000,000       | 1,000,000      | 06.32  | <video controls autoplay loop src="https://cdn-uploads.huggingface.co/production/uploads/633c1daf31c06121a58f2df9/_hDMocxjmzutc0t5FfoTX.mp4"></video>  |
| maze        | 9,000,000       | 1,000,000      | 09.95  | <video controls autoplay loop src="https://cdn-uploads.huggingface.co/production/uploads/633c1daf31c06121a58f2df9/uhNdDPuNhZpxVns91Ba-9.mp4"></video>  |
| miner       | 9,000,000       | 1,000,000      | 12.02  | <video controls autoplay loop src="https://cdn-uploads.huggingface.co/production/uploads/633c1daf31c06121a58f2df9/ElpJ8l2WHJGrprZ3-giHU.mp4"></video>  |
| ninja       | 9,000,000       | 1,000,000      | 09.32  | <video controls autoplay loop src="https://cdn-uploads.huggingface.co/production/uploads/633c1daf31c06121a58f2df9/b9i-fb2Twh8XmBBNf2DRG.mp4"></video>  |
| plunder     | 9,000,000       | 1,000,000      | 24.18  | <video controls autoplay loop src="https://cdn-uploads.huggingface.co/production/uploads/633c1daf31c06121a58f2df9/JPeGNOVzrotuYUjfzZj40.mp4"></video>  |
| starpilot   | 9,000,000       | 1,000,000      | 49.79  | <video controls autoplay loop src="https://cdn-uploads.huggingface.co/production/uploads/633c1daf31c06121a58f2df9/wY9lZgkw5tor19hCWmm6A.mp4"></video>  |


## Dataset Structure
### Data Instances
Each data instance represents a single step consisting of tuples of the form (observation, action, reward, done, truncated) = (o_t, a_t, r_t, terminated_t, truncated_t).

```json
{'action': 1,
 'observation': [[[0, 166, 253],
                  [0, 174, 255],
                  [0, 170, 251],
                  [0, 191, 255],
                  [0, 191, 255],
                  [0, 221, 255],
                  [0, 243, 255],
                  [0, 248, 255],
                  [0, 243, 255],
                  [10, 239, 255],
                  [25, 255, 255],
                  [0, 241, 255],
                  [0, 235, 255],
                  [17, 240, 255],
                  [10, 243, 255],
                  [27, 253, 255],
                  [39, 255, 255],
                  [58, 255, 255],
                  [85, 255, 255],
                  [111, 255, 255],
                  [135, 255, 255],
                  [151, 255, 255],
                  [173, 255, 255],
...
                  [0, 0, 37],
                  [0, 0, 39]]],
 'reward': 0.0,
 'terminated': False,
 'truncated': False}
```

### Data Fields
- `observation`: The current RGB observation from the environment.
- `action`: The action predicted by the agent for the current observation.
- `reward`: The received reward for the current observation.
- `terminated`: If the episode has terminated with the current observation.
- `truncated`: If the episode is truncated with the current observation.

### Data Splits
The dataset is divided into a `train` (90%) and `test` (10%) split. Each environment-dataset has in sum 10M steps (data points).

## Dataset Creation
The dataset was created by training an RL agent with [PPO](https://arxiv.org/abs/1707.06347) for 25M steps in each environment. The trajectories where generated by sampling from the predicted action distribution at each step (not taking the argmax). The environments were created on `distribution_mode=easy` and with unlimited levels.

## Procgen Benchmark
The [Procgen Benchmark](https://openai.com/index/procgen-benchmark/), released by OpenAI, consists of 16 procedurally-generated environments designed to measure how quickly reinforcement learning (RL) agents learn generalizable skills. It emphasizes experimental convenience, high diversity within and across environments, and is ideal for evaluating both sample efficiency and generalization. The benchmark allows for distinct training and test sets in each environment, making it a standard research platform for the OpenAI RL team. It aims to address the need for more diverse RL benchmarks compared to complex environments like Dota and StarCraft.