Procgen Integration (#197)

Procgen library change in https://github.com/Trinkle23897/procgen - Fix Qt library include error - Fix globalGameRegistry == nullptr issue by adding another explicit make_* function (thanks to Yukun!) - Disable auto-reset because envpool has already handled this feature Changes: - Add new dependency `qtdeclarative5-dev`, rebuild docker image - Update pypi wheel from manylinux_2_17_x86_64 to manylinux_2_24_x86_64 because of libstdc++.so.6 symbol - Add 42 procgen envs - Refactor `base_path` Co-authored-by: Jiayi Weng <[email protected]> Co-authored-by: Leo Guo <[email protected]>
sail-sg · Jan 3, 2023 · d44c541 · d44c541
1 parent 60459ad
commit d44c541
Show file tree

Hide file tree

Showing 34 changed files with 818 additions and 40 deletions.
diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
@@ -9,7 +9,7 @@ on:
 jobs:
   release:
     runs-on: ubuntu-latest
-    container: trinkle23897/envpool-release:2022-12-29-03c3d95
+    container: trinkle23897/envpool-release:2023-01-02-5f1a5fd
     strategy:
       matrix:
         python-version: ['3.7', '3.8', '3.9', '3.10', '3.11']
@@ -27,7 +27,7 @@ jobs:
     - name: Build
       run: |
         make pypi-wheel
-        pip3 install wheelhouse/*.whl --force-reinstall
+        pip3 install dist/*.whl --force-reinstall
     - name: Test
       run: |
         make release-test

diff --git a/Makefile b/Makefile
@@ -180,7 +180,7 @@ docker-release-launch: docker-release
 	docker run --network=host -v /:/host --shm-size=4gb -it $(PROJECT_NAME)-release:$(DOCKER_TAG) bash
 
 pypi-wheel: auditwheel-install bazel-release
-	ls dist/*.whl -Art | tail -n 1 | xargs auditwheel repair --plat manylinux_2_17_x86_64
+	ls dist/*.whl -Art | tail -n 1 | xargs auditwheel repair --plat manylinux_2_24_x86_64
 
 release-test1:
 	cd envpool && python3 make_test.py

diff --git a/README.md b/README.md
@@ -16,7 +16,7 @@
 - [x] [ViZDoom single player](https://envpool.readthedocs.io/en/latest/env/vizdoom.html)
 - [x] [DeepMind Control Suite](https://envpool.readthedocs.io/en/latest/env/dm_control.html)
 - [x] [Box2D](https://envpool.readthedocs.io/en/latest/env/box2d.html)
-- [ ] Procgen
+- [x] [Procgen](https://envpool.readthedocs.io/en/latest/env/procgen.html)
 - [ ] Minigrid
 
 Here are EnvPool's several highlights:

diff --git a/WORKSPACE b/WORKSPACE
@@ -8,6 +8,20 @@ load("//envpool:workspace1.bzl", workspace1 = "workspace")
 
 workspace1()
 
+# QT special, cannot move to workspace2.bzl, not sure why
+
+load("@local_config_qt//:local_qt.bzl", "local_qt_path")
+
+new_local_repository(
+    name = "qt",
+    build_file = "@com_justbuchanan_rules_qt//:qt.BUILD",
+    path = local_qt_path(),
+)
+
+load("@com_justbuchanan_rules_qt//tools:qt_toolchain.bzl", "register_qt_toolchains")
+
+register_qt_toolchains()
+
 load("//envpool:pip.bzl", pip_workspace = "workspace")
 
 pip_workspace()
diff --git a/docker/dev-cn.dockerfile b/docker/dev-cn.dockerfile
@@ -7,7 +7,7 @@ ARG HOME=/root
 ARG PATH=$PATH:$HOME/go/bin
 
 RUN apt-get update \
-    && apt-get install -y python3-pip python3-dev golang-1.18 clang-format-11 git wget swig tmux clang-tidy vim \
+    && apt-get install -y python3-pip python3-dev golang-1.18 clang-format-11 git wget swig tmux clang-tidy vim qtdeclarative5-dev \
     && rm -rf /var/lib/apt/lists/*
 RUN ln -s /usr/bin/python3 /usr/bin/python
 RUN ln -sf /usr/lib/go-1.18/bin/go /usr/bin/go

diff --git a/docker/dev.dockerfile b/docker/dev.dockerfile
@@ -7,7 +7,7 @@ ARG HOME=/root
 ARG PATH=$PATH:$HOME/go/bin
 
 RUN apt-get update \
-    && apt-get install -y python3-pip python3-dev golang-1.18 clang-format-11 git wget swig tmux clang-tidy vim \
+    && apt-get install -y python3-pip python3-dev golang-1.18 clang-format-11 git wget swig tmux clang-tidy vim qtdeclarative5-dev \
     && rm -rf /var/lib/apt/lists/*
 RUN ln -s /usr/bin/python3 /usr/bin/python
 RUN ln -sf /usr/lib/go-1.18/bin/go /usr/bin/go

diff --git a/docker/release.dockerfile b/docker/release.dockerfile
@@ -14,7 +14,8 @@ RUN apt-get update && apt-get install -y software-properties-common && add-apt-r
 RUN apt-get update \
     && apt-get install -y git curl wget gcc-9 g++-9 build-essential swig make \
     zlib1g-dev libbz2-dev libreadline-dev libsqlite3-dev libncursesw5-dev libffi-dev liblzma-dev \
-    llvm xz-utils tk-dev libxml2-dev libxmlsec1-dev
+    llvm xz-utils tk-dev libxml2-dev libxmlsec1-dev qtdeclarative5-dev \
+    && rm -rf /var/lib/apt/lists/*
 # use self-compiled openssl instead of system provided (1.0.2)
 RUN apt-get remove -y libssl-dev
 

diff --git a/docs/env/procgen.rst b/docs/env/procgen.rst
@@ -0,0 +1,97 @@
+Procgen
+=======
+
+We use ``procgen==0.10.7`` as the codebase.
+See https://github.com/openai/procgen/tree/0.10.7
+
+
+Options
+-------
+
+* ``task_id (str)``: see available tasks below;
+* ``num_envs (int)``: how many environments you would like to create;
+* ``batch_size (int)``: the expected batch size for return result, default to
+  ``num_envs``;
+* ``num_threads (int)``: the maximum thread number for executing the actual
+  ``env.step``, default to ``batch_size``;
+* ``seed (int)``: the environment seed, default to ``42``;
+* ``max_episode_steps (int)``: the maximum number of steps for one episode,
+  each procgen game has different timeout value;
+* ``env_name (str)``: one of 16 procgen env name;
+* ``num_levels (int)``: default to ``0``;
+* ``start_level (int)``: default to ``0``;
+* ``use_sequential_levels (bool)``: default to ``False``;
+* ``center_agent (bool)``: default to ``True``;
+* ``use_backgrounds (bool)``: default to ``True``;
+* ``use_monochrome_assets (bool)``: default to ``False``;
+* ``restrict_themes (bool)``: default to ``False``;
+* ``use_generated_assets (bool)``: default to ``False``;
+* ``paint_vel_info (bool)``: default to ``False``;
+* ``use_easy_jump (bool)``: default to ``False``;
+* ``distribution_mode (int)``: one of ``(0, 1, 2, 10)``; ``0`` stands for easy
+  mode, ``1`` stands for hard mode, ``2`` stands for extreme mode, ``10``
+  stands for memory mode. The default value is determined by ``task_id``.
+
+Note: arguments after ``env_name`` are provided by procgen environment itself.
+We keep the default value as-is. We haven't tested the setting of
+``use_sequential_levels == True``, and have no promise it is aligned with the
+original version of procgen (PRs for fixing this issue are highly welcome).
+
+
+Observation Space
+-----------------
+
+The observation image size is ``(64, 64, 3)``.
+
+
+Action Space
+------------
+
+15 action buttons in total, ranging from 0 to 14.
+
+
+Available Tasks
+---------------
+
+* ``BigfishEasy-v0``
+* ``BigfishHard-v0``
+* ``BossfightEasy-v0``
+* ``BossfightHard-v0``
+* ``CaveflyerEasy-v0``
+* ``CaveflyerHard-v0``
+* ``CaveflyerMemory-v0``
+* ``ChaserEasy-v0``
+* ``ChaserHard-v0``
+* ``ChaserExtreme-v0``
+* ``ClimberEasy-v0``
+* ``ClimberHard-v0``
+* ``CoinrunEasy-v0``
+* ``CoinrunHard-v0``
+* ``DodgeballEasy-v0``
+* ``DodgeballHard-v0``
+* ``DodgeballExtreme-v0``
+* ``DodgeballMemory-v0``
+* ``FruitbotEasy-v0``
+* ``FruitbotHard-v0``
+* ``HeistEasy-v0``
+* ``HeistHard-v0``
+* ``HeistMemory-v0``
+* ``JumperEasy-v0``
+* ``JumperHard-v0``
+* ``JumperMemory-v0``
+* ``LeaperEasy-v0``
+* ``LeaperHard-v0``
+* ``LeaperExtreme-v0``
+* ``MazeEasy-v0``
+* ``MazeHard-v0``
+* ``MazeMemory-v0``
+* ``MinerEasy-v0``
+* ``MinerHard-v0``
+* ``MinerMemory-v0``
+* ``NinjaEasy-v0``
+* ``NinjaHard-v0``
+* ``PlunderEasy-v0``
+* ``PlunderHard-v0``
+* ``StarpilotEasy-v0``
+* ``StarpilotHard-v0``
+* ``StarpilotExtreme-v0``
diff --git a/docs/index.rst b/docs/index.rst
@@ -91,6 +91,7 @@ stable version through `envpool.readthedocs.io/en/stable/
    env/classic_control
    env/dm_control
    env/mujoco_gym
+   env/procgen
    env/toy_text
    env/vizdoom
 

diff --git a/docs/spelling_wordlist.txt b/docs/spelling_wordlist.txt
@@ -68,3 +68,4 @@ api
 jit
 mins
 lidar
+procgen
diff --git a/envpool/BUILD b/envpool/BUILD
@@ -35,6 +35,7 @@ py_library(
         "//envpool/classic_control:classic_control_registration",
         "//envpool/mujoco:mujoco_dmc_registration",
         "//envpool/mujoco:mujoco_gym_registration",
+        "//envpool/procgen:procgen_registration",
         "//envpool/toy_text:toy_text_registration",
         "//envpool/vizdoom:vizdoom_registration",
     ],
@@ -51,6 +52,7 @@ py_library(
         "//envpool/classic_control",
         "//envpool/mujoco:mujoco_dmc",
         "//envpool/mujoco:mujoco_gym",
+        "//envpool/procgen",
         "//envpool/python",
         "//envpool/toy_text",
         "//envpool/vizdoom",

diff --git a/envpool/__init__.py b/envpool/__init__.py
@@ -24,7 +24,7 @@
   register,
 )
 
-__version__ = "0.7.1"
+__version__ = "0.8.0"
 __all__ = [
   "register",
   "make",

diff --git a/envpool/atari/registration.py b/envpool/atari/registration.py
@@ -15,9 +15,7 @@
 
 import os
 
-from envpool.registration import register
-
-base_path = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
+from envpool.registration import base_path, register
 
 atari_rom_path = os.path.join(base_path, "atari", "roms")
 atari_game_list = sorted(
@@ -34,6 +32,5 @@
     gym_cls="AtariGymEnvPool",
     gymnasium_cls="AtariGymnasiumEnvPool",
     task=game,
-    base_path=base_path,
     max_episode_steps=108000,
   )
diff --git a/envpool/entry.py b/envpool/entry.py
@@ -18,5 +18,6 @@
 import envpool.classic_control.registration  # noqa: F401
 import envpool.mujoco.dmc.registration  # noqa: F401
 import envpool.mujoco.gym.registration  # noqa: F401
+import envpool.procgen.registration  # noqa: F401
 import envpool.toy_text.registration  # noqa: F401
 import envpool.vizdoom.registration  # noqa: F401
diff --git a/envpool/make_test.py b/envpool/make_test.py
@@ -194,6 +194,54 @@ def test_make_mujoco_dmc(self) -> None:
       ]
     )
 
+  def test_make_procgen(self) -> None:
+    self.check_step(
+      [
+        "BigfishEasy-v0",
+        "BigfishHard-v0",
+        "BossfightEasy-v0",
+        "BossfightHard-v0",
+        "CaveflyerEasy-v0",
+        "CaveflyerHard-v0",
+        "CaveflyerMemory-v0",
+        "ChaserEasy-v0",
+        "ChaserHard-v0",
+        "ChaserExtreme-v0",
+        "ClimberEasy-v0",
+        "ClimberHard-v0",
+        "CoinrunEasy-v0",
+        "CoinrunHard-v0",
+        "DodgeballEasy-v0",
+        "DodgeballHard-v0",
+        "DodgeballExtreme-v0",
+        "DodgeballMemory-v0",
+        "FruitbotEasy-v0",
+        "FruitbotHard-v0",
+        "HeistEasy-v0",
+        "HeistHard-v0",
+        "HeistMemory-v0",
+        "JumperEasy-v0",
+        "JumperHard-v0",
+        "JumperMemory-v0",
+        "LeaperEasy-v0",
+        "LeaperHard-v0",
+        "LeaperExtreme-v0",
+        "MazeEasy-v0",
+        "MazeHard-v0",
+        "MazeMemory-v0",
+        "MinerEasy-v0",
+        "MinerHard-v0",
+        "MinerMemory-v0",
+        "NinjaEasy-v0",
+        "NinjaHard-v0",
+        "PlunderEasy-v0",
+        "PlunderHard-v0",
+        "StarpilotEasy-v0",
+        "StarpilotHard-v0",
+        "StarpilotExtreme-v0",
+      ]
+    )
+
 
 if __name__ == "__main__":
   absltest.main()
diff --git a/envpool/mujoco/dmc/registration.py b/envpool/mujoco/dmc/registration.py
@@ -13,14 +13,8 @@
 # limitations under the License.
 """Mujoco dm_control suite env registration."""
 
-import os
-
 from envpool.registration import register
 
-base_path = os.path.abspath(
-  os.path.join(os.path.dirname(__file__), "..", "..")
-)
-
 # from suite.BENCHMARKING
 dmc_mujoco_envs = [
   ("acrobot", "swingup", 1000),
@@ -72,7 +66,6 @@
     dm_cls=f"Dmc{domain_name}DMEnvPool",
     gym_cls=f"Dmc{domain_name}GymEnvPool",
     gymnasium_cls=f"Dmc{domain_name}GymnasiumEnvPool",
-    base_path=base_path,
     task_name=task,
     max_episode_steps=max_episode_steps,
   )
diff --git a/envpool/mujoco/gym/registration.py b/envpool/mujoco/gym/registration.py
@@ -13,14 +13,8 @@
 # limitations under the License.
 """Mujoco gym env registration."""
 
-import os
-
 from envpool.registration import register
 
-base_path = os.path.abspath(
-  os.path.join(os.path.dirname(__file__), "..", "..")
-)
-
 gym_mujoco_envs = [
   ("Ant", "v3", False, 1000),
   ("Ant", "v4", True, 1000),
@@ -57,7 +51,6 @@
     dm_cls=f"Gym{task}DMEnvPool",
     gym_cls=f"Gym{task}GymEnvPool",
     gymnasium_cls=f"Gym{task}GymnasiumEnvPool",
-    base_path=base_path,
     post_constraint=post_constraint,
     max_episode_steps=max_episode_steps,
     **extra_args,