From 2fbd37e0391de596ae1427e49d0e99ae49a8e018 Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Tue, 9 Apr 2024 09:10:40 +0100
Subject: [PATCH 01/17] Add new python versions

---
 .github/workflows/test_and_deploy.yml | 2 +-
 .github/workflows/test_package.yml    | 2 +-
 .gitignore                            | 4 ++++
 .pre-commit-config.yaml               | 6 +++---
 4 files changed, 9 insertions(+), 5 deletions(-)

diff --git a/.github/workflows/test_and_deploy.yml b/.github/workflows/test_and_deploy.yml
index 2812c2abb..eafa829da 100644
--- a/.github/workflows/test_and_deploy.yml
+++ b/.github/workflows/test_and_deploy.yml
@@ -18,7 +18,7 @@ jobs:
     runs-on: ${{ matrix.os }}
     strategy:
       matrix:
-        python-version: ["3.6", "3.7", "3.8", "3.9"]
+        python-version: ["3.8", "3.9", "3.10", "3.11"]
         os: [ubuntu-20.04, macos-latest]
       fail-fast: false
     steps:
diff --git a/.github/workflows/test_package.yml b/.github/workflows/test_package.yml
index 0c8c07ffb..f501f7a42 100644
--- a/.github/workflows/test_package.yml
+++ b/.github/workflows/test_package.yml
@@ -11,7 +11,7 @@ jobs:
     runs-on: macos-latest
     strategy:
       matrix:
-        python-version: ["3.5", "3.6", "3.7", "3.8"]
+        python-version: ["3.8", "3.9", "3.10", "3.11"]
       fail-fast: false
     steps:
     - name: Setup Python ${{ matrix.python-version }} env
diff --git a/.gitignore b/.gitignore
index 9d4ea2863..e7dd6fe85 100644
--- a/.gitignore
+++ b/.gitignore
@@ -181,6 +181,7 @@ venv.bak/
 
 # IDE
 .idea/
+.vscode/
 
 # Rope project settings
 .ropeproject
@@ -205,3 +206,6 @@ nle/version.py
 nle_data/
 nle/fbs/
 nle/nethackdir
+
+# Generated during tests
+nle.ttyrec3.bz2
\ No newline at end of file
diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index a6ffdd9bc..20bfce297 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -1,11 +1,11 @@
 repos:
 -   repo: https://github.com/ambv/black
-    rev: 22.3.0
+    rev: 24.3.0
     hooks:
     - id: black
       language_version: python3.8
 -   repo: https://github.com/pycqa/flake8
-    rev: '3.9.2'
+    rev: '7.0.0'
     hooks:
     - id: flake8
       additional_dependencies: [flake8-bugbear]
@@ -23,7 +23,7 @@ repos:
         language: system
         files: ^(src\/nle|include\/nle|win\/rl|sys\/unix\/nle).*\.(c|cc|cxx|cpp|cu|h|hpp|hxx|cuh|proto)$
 -   repo: https://github.com/pycqa/isort
-    rev: 5.8.0
+    rev: 5.13.2
     hooks:
       - id: isort
         name: isort (python)

From 8ab7e928670942702e2f4d05f5a3885a986f2617 Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Tue, 9 Apr 2024 12:21:02 +0100
Subject: [PATCH 02/17] Add isort to dev dependencies

---
 setup.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/setup.py b/setup.py
index 2eda2828e..f2862f1f6 100644
--- a/setup.py
+++ b/setup.py
@@ -102,6 +102,7 @@ def build_extension(self, ext):
 extras_deps = {
     "dev": [
         "pre-commit>=2.0.1",
+        "isort>=5.13.2",
         "black>=19.10b0",
         "cmake_format>=0.6.10",
         "flake8>=3.7",

From e5a0174bf16853d378182727bd8361a1848ea9df Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Tue, 9 Apr 2024 12:23:33 +0100
Subject: [PATCH 03/17] Switch GitHub Action PyPI publish dependency to
 release/v1

---
 .github/workflows/test_and_deploy.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/workflows/test_and_deploy.yml b/.github/workflows/test_and_deploy.yml
index eafa829da..0c02c31a3 100644
--- a/.github/workflows/test_and_deploy.yml
+++ b/.github/workflows/test_and_deploy.yml
@@ -138,7 +138,7 @@ jobs:
     # NOTE: we assume that dist/ contains a built sdist (and only that).
     #       Yes, we could be more defensively, but What Could Go Wrong?™
     - name: Publish package to PyPI
-      uses: pypa/gh-action-pypi-publish@master
+      uses: pypa/gh-action-pypi-publish@release/v1
       with:
         user: __token__
         password: ${{ secrets.PYPI_TOKEN }}

From 1e5eb059409ec2846c83286a1949ba9be6907680 Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Tue, 9 Apr 2024 12:54:48 +0100
Subject: [PATCH 04/17] Fix Black code format complaint

---
 nle/scripts/collect_env.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/nle/scripts/collect_env.py b/nle/scripts/collect_env.py
index 9fb5ed67d..41b842cbe 100644
--- a/nle/scripts/collect_env.py
+++ b/nle/scripts/collect_env.py
@@ -293,9 +293,9 @@ def get_env_info():
         cuda_available_str = torch.cuda.is_available()
         cuda_version_str = torch.version.cuda
     else:
-        torch_version_str = (
+        torch_version_str = cuda_available_str = cuda_version_str = (
             torch_debug_mode_str
-        ) = cuda_available_str = cuda_version_str = "N/A"
+        ) = "N/A"
 
     return SystemEnv(
         nle_version=nle_version,

From 990b8eb0eb5aa5a979fb26e0def8a6721cee8ade Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Wed, 10 Apr 2024 18:49:58 +0100
Subject: [PATCH 05/17] Move distribution package to Python 3.11

---
 .github/workflows/test_and_deploy.yml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/.github/workflows/test_and_deploy.yml b/.github/workflows/test_and_deploy.yml
index 0c02c31a3..c70b45061 100644
--- a/.github/workflows/test_and_deploy.yml
+++ b/.github/workflows/test_and_deploy.yml
@@ -62,14 +62,14 @@ jobs:
         path: nle_test_ci_${{ github.sha }}.tar.gz
 
   test_sdist:
-    name: Test sdist on MacOS w/ Py3.8
+    name: Test sdist on MacOS w/ Py3.11
     needs: test_repo
     runs-on: macos-latest
     steps:
-    - name: Setup Python 3.8 env
+    - name: Setup Python 3.11 env
       uses: actions/setup-python@v2
       with:
-        python-version: 3.8
+        python-version: 3.11
     - name: Ensure latest pip & wheel
       run: "python -m pip install -q --upgrade pip wheel"
     - name: Install dependencies

From 7656335d0397e9c108384bced6ad3aa95d1d235e Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Wed, 17 Apr 2024 13:11:21 +0100
Subject: [PATCH 06/17] Upgrade to Gym 26.2 and change render() calls for API
 compatibility

---
 nle/env/base.py        | 7 +++++--
 nle/scripts/play.py    | 3 ++-
 nle/tests/test_envs.py | 4 ++--
 setup.py               | 2 +-
 4 files changed, 10 insertions(+), 6 deletions(-)

diff --git a/nle/env/base.py b/nle/env/base.py
index 2f3a71ec4..6f813c000 100644
--- a/nle/env/base.py
+++ b/nle/env/base.py
@@ -193,6 +193,7 @@ def __init__(
         allow_all_yn_questions=False,
         allow_all_modes=False,
         spawn_monsters=True,
+        render_mode=str | None,
     ):
         """Constructs a new NLE environment.
 
@@ -230,6 +231,7 @@ def __init__(
         self._allow_all_yn_questions = allow_all_yn_questions
         self._allow_all_modes = allow_all_modes
         self._save_ttyrec_every = save_ttyrec_every
+        self.render_mode = render_mode
 
         if actions is None:
             actions = FULL_ACTIONS
@@ -475,8 +477,9 @@ def get_seeds(self):
         """
         return self.nethack.get_current_seeds()
 
-    def render(self, mode="human"):
+    def render(self):
         """Renders the state of the environment."""
+        mode = self.render_mode
 
         if mode == "human":
             obs = self.last_observation
@@ -515,7 +518,7 @@ def render(self, mode="human"):
             # TODO: Why return a string here but print in the other branches?
             return "\n".join([line.tobytes().decode("utf-8") for line in chars])
 
-        return super().render(mode=mode)
+        return super().render()
 
     def __repr__(self):
         return "<%s>" % self.__class__.__name__
diff --git a/nle/scripts/play.py b/nle/scripts/play.py
index 141d02eca..f8b3f354e 100644
--- a/nle/scripts/play.py
+++ b/nle/scripts/play.py
@@ -92,6 +92,7 @@ def play():
             allow_all_yn_questions=True,
             allow_all_modes=True,
             wizard=FLAGS.wizard,
+            render_mode=FLAGS.render_mode,
         )
         if FLAGS.seeds is not None:
             env.seed(FLAGS.seeds)
@@ -117,7 +118,7 @@ def play():
                 act_str = repr(env.actions[action]) if action is not None else ""
                 print(f"Previous action: {str(act_str):64s}")
                 print("-" * 8)
-                env.render(FLAGS.render_mode)
+                env.render()
                 print("-" * 8)
                 print(obs["blstats"])
                 if not FLAGS.print_frames_separately:
diff --git a/nle/tests/test_envs.py b/nle/tests/test_envs.py
index 00704c8e6..682843aa2 100644
--- a/nle/tests/test_envs.py
+++ b/nle/tests/test_envs.py
@@ -331,14 +331,14 @@ def test_seed_rollout_seeded_int(self, env_name, rollout_len):
         compare_rollouts(env0, env1, rollout_len)
 
     def test_render_ansi(self, env_name, rollout_len):
-        env = gym.make(env_name)
+        env = gym.make(env_name, render_mode="ansi")
         env.reset()
         for _ in range(rollout_len):
             action = env.action_space.sample()
             _, _, done, _ = env.step(action)
             if done:
                 env.reset()
-            output = env.render(mode="ansi")
+            output = env.render()
             assert isinstance(output, str)
             assert len(output.replace("\n", "")) == np.prod(nle.env.DUNGEON_SHAPE)
 
diff --git a/setup.py b/setup.py
index f2862f1f6..f48360a86 100644
--- a/setup.py
+++ b/setup.py
@@ -163,7 +163,7 @@ def build_extension(self, ext):
         ext_modules=[setuptools.Extension("nle", sources=[])],
         cmdclass={"build_ext": CMakeBuild},
         setup_requires=["pybind11>=2.2"],
-        install_requires=["pybind11>=2.2", "numpy>=1.16", "gym>=0.15,<=0.23"],
+        install_requires=["pybind11>=2.2", "numpy>=1.16", "gym>=0.15"],
         extras_require=extras_deps,
         python_requires=">=3.5",
         classifiers=[

From 861e72b5c4915473e656a168dfd6bf2f9ba22164 Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Wed, 17 Apr 2024 15:33:42 +0100
Subject: [PATCH 07/17] Environment registry changed from class to dict -
 changing access

---
 nle/tests/test_envs.py | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/nle/tests/test_envs.py b/nle/tests/test_envs.py
index 682843aa2..3e849a22b 100644
--- a/nle/tests/test_envs.py
+++ b/nle/tests/test_envs.py
@@ -17,14 +17,12 @@
 
 
 def get_nethack_env_ids():
-    specs = gym.envs.registry.all()
+    specs = gym.envs.registry.keys()
     # Ignoring base environment, since we can't handle random actions yet with
     # the full action space, and this requires a whole different set of tests.
     # For now this is OK, since NetHackScore-v0 is very similar.
     return [
-        spec.id
-        for spec in specs
-        if spec.id.startswith("NetHack") and spec.id != "NetHack-v0"
+        spec for spec in specs if spec.startswith("NetHack") and spec != "NetHack-v0"
     ]
 
 

From b613742f229f3882a0b67ffa180971ec8c6c5d2d Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Wed, 17 Apr 2024 19:03:37 +0100
Subject: [PATCH 08/17] Change reset() return values for API compliance

---
 nle/env/base.py | 35 ++++++++++++++++++++++++-----------
 1 file changed, 24 insertions(+), 11 deletions(-)

diff --git a/nle/env/base.py b/nle/env/base.py
index 6f813c000..824a11a41 100644
--- a/nle/env/base.py
+++ b/nle/env/base.py
@@ -331,6 +331,19 @@ def _get_observation(self, observation):
             for key, i in zip(self._original_observation_keys, self._original_indices)
         }
 
+    def _get_end_status(self, observation, done):
+        if self._check_abort(observation):
+            end_status = self.StepStatus.ABORTED
+        else:
+            end_status = self._is_episode_end(observation)
+        return self.StepStatus(done or end_status)
+
+    def _get_information(self, end_status):
+        info = {}
+        info["end_status"] = end_status
+        info["is_ascended"] = self.nethack.how_done() == nethack.ASCENDED
+        return info
+
     def print_action_meanings(self):
         for a_idx, a in enumerate(self.actions):
             print(a_idx, a)
@@ -369,11 +382,7 @@ def step(self, action: int):
 
         self.last_observation = observation
 
-        if self._check_abort(observation):
-            end_status = self.StepStatus.ABORTED
-        else:
-            end_status = self._is_episode_end(observation)
-        end_status = self.StepStatus(done or end_status)
+        end_status = self._get_end_status(observation, done)
 
         reward = float(
             self._reward_fn(last_observation, action, observation, end_status)
@@ -384,11 +393,12 @@ def step(self, action: int):
             self._quit_game(observation, done)
             done = True
 
-        info = {}
-        info["end_status"] = end_status
-        info["is_ascended"] = self.nethack.how_done() == nethack.ASCENDED
-
-        return self._get_observation(observation), reward, done, info
+        return (
+            self._get_observation(observation),
+            reward,
+            done,
+            self._get_information(end_status),
+        )
 
     def _in_moveloop(self, observation):
         program_state = observation[self._program_state_index]
@@ -416,6 +426,7 @@ def reset(self, wizkit_items=None):
         )
 
         self._steps = 0
+        done = False
 
         for _ in range(1000):
             # Get past initial phase of game. This should make sure
@@ -434,7 +445,9 @@ def reset(self, wizkit_items=None):
             )
             return self.reset(wizkit_items=wizkit_items)
 
-        return self._get_observation(self.last_observation)
+        return self._get_observation(self.last_observation), self._get_information(
+            self._get_end_status(self.last_observation, done)
+        )
 
     def close(self):
         self._close_nethack()

From b309918d8f169d608f6b7ff009c7cd4e4b41c330 Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Wed, 17 Apr 2024 19:25:24 +0100
Subject: [PATCH 09/17] Add additional parameters to reset() for API compliance

---
 nle/env/base.py | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/nle/env/base.py b/nle/env/base.py
index 824a11a41..202406bdc 100644
--- a/nle/env/base.py
+++ b/nle/env/base.py
@@ -404,7 +404,7 @@ def _in_moveloop(self, observation):
         program_state = observation[self._program_state_index]
         return program_state[3]  # in_moveloop
 
-    def reset(self, wizkit_items=None):
+    def reset(self, seed=None, options=None, wizkit_items=None):
         """Resets the environment.
 
         Note:
@@ -413,9 +413,12 @@ def reset(self, wizkit_items=None):
             fail in case Nethack is initialized with some uncommon options.
 
         Returns:
-            [dict] Observation of the state as defined by
-                `self.observation_space`.
+            (tuple) (Observation of the state as defined by
+                `self.observation_space`,
+                Extra information)
         """
+        super().reset(seed=seed)
+
         self._episode += 1
         if self.savedir and self._episode % self._save_ttyrec_every == 0:
             new_ttyrec = self._ttyrec_pattern % self._episode

From e7c4ea2273ebe78d81513497f44e9019b3b6462f Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Fri, 19 Apr 2024 16:36:43 +0100
Subject: [PATCH 10/17] Fix to rendering to match gym passive checkers

---
 nle/env/base.py | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/nle/env/base.py b/nle/env/base.py
index 202406bdc..c0f724314 100644
--- a/nle/env/base.py
+++ b/nle/env/base.py
@@ -147,7 +147,7 @@ class NLE(gym.Env):
         >>> env.render()
     """
 
-    metadata = {"render.modes": ["human", "ansi"]}
+    metadata = {"render_modes": ["human", "ansi", "full"]}
 
     class StepStatus(enum.IntEnum):
         """Specifies the status of the terminal state.
@@ -193,7 +193,7 @@ def __init__(
         allow_all_yn_questions=False,
         allow_all_modes=False,
         spawn_monsters=True,
-        render_mode=str | None,
+        render_mode="human",
     ):
         """Constructs a new NLE environment.
 
@@ -225,6 +225,9 @@ def __init__(
                 If set to False, only skip click through 'MORE' on death.
             spawn_monsters: If False, disables normal NetHack behavior to randomly
                 create monsters.
+            render_mode (str): mode used to render the screen. One of
+                "human" | "ansi" | "full".
+                Defaults to "human", i.e. what a human would see playing the game.
         """
         self.character = character
         self._max_episode_steps = max_episode_steps
@@ -534,8 +537,6 @@ def render(self):
             # TODO: Why return a string here but print in the other branches?
             return "\n".join([line.tobytes().decode("utf-8") for line in chars])
 
-        return super().render()
-
     def __repr__(self):
         return "<%s>" % self.__class__.__name__
 

From 0d662e4b7027d2753235406cc28a14c26b4faec8 Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Fri, 19 Apr 2024 17:13:06 +0100
Subject: [PATCH 11/17] Update calls to reset() for new API

---
 nle/agent/agent.py     |  8 ++++----
 nle/env/base.py        | 10 +++++++---
 nle/scripts/play.py    |  2 +-
 nle/tests/test_envs.py | 12 ++++++------
 4 files changed, 18 insertions(+), 14 deletions(-)

diff --git a/nle/agent/agent.py b/nle/agent/agent.py
index 208126884..6498d11cf 100644
--- a/nle/agent/agent.py
+++ b/nle/agent/agent.py
@@ -141,7 +141,7 @@ def compute_policy_gradient_loss(logits, actions, advantages):
 
 
 def create_env(name, *args, **kwargs):
-    return gym.make(name, observation_keys=("glyphs", "blstats"), *args, **kwargs)
+    return gym.make(name, *args, observation_keys=("glyphs", "blstats"), **kwargs)
 
 
 def act(
@@ -350,8 +350,8 @@ def initial(self):
         self.episode_return = torch.zeros(1, 1)
         self.episode_step = torch.zeros(1, 1, dtype=torch.int32)
         initial_done = torch.ones(1, 1, dtype=torch.uint8)
-
-        result = _format_observations(self.gym_env.reset())
+        obs, reset_info = self.gym_env.reset()
+        result = _format_observations(obs)
         result.update(
             reward=initial_reward,
             done=initial_done,
@@ -368,7 +368,7 @@ def step(self, action):
         episode_step = self.episode_step
         episode_return = self.episode_return
         if done:
-            observation = self.gym_env.reset()
+            observation, reset_info = self.gym_env.reset()
             self.episode_return = torch.zeros(1, 1)
             self.episode_step = torch.zeros(1, 1, dtype=torch.int32)
 
diff --git a/nle/env/base.py b/nle/env/base.py
index c0f724314..d29d135b0 100644
--- a/nle/env/base.py
+++ b/nle/env/base.py
@@ -142,12 +142,16 @@ class NLE(gym.Env):
 
     Examples:
         >>> env = NLE()
-        >>> obs = env.reset()
+        >>> obs, reset_info = env.reset()
         >>> obs, reward, done, info = env.step(0)
         >>> env.render()
     """
 
-    metadata = {"render_modes": ["human", "ansi", "full"]}
+    # Gym expects an fps rate > 0 for render checks, but
+    # NetHack doesn't have any. Setting it to 42 because
+    # that's always the answer to life, the universe and
+    # everything.
+    metadata = {"render_modes": ["human", "ansi", "full"], "render_fps": 42}
 
     class StepStatus(enum.IntEnum):
         """Specifies the status of the terminal state.
@@ -418,7 +422,7 @@ def reset(self, seed=None, options=None, wizkit_items=None):
         Returns:
             (tuple) (Observation of the state as defined by
                 `self.observation_space`,
-                Extra information)
+                Extra game state information)
         """
         super().reset(seed=seed)
 
diff --git a/nle/scripts/play.py b/nle/scripts/play.py
index f8b3f354e..5d4b5f5d1 100644
--- a/nle/scripts/play.py
+++ b/nle/scripts/play.py
@@ -97,7 +97,7 @@ def play():
         if FLAGS.seeds is not None:
             env.seed(FLAGS.seeds)
 
-    obs = env.reset()
+    obs, reset_info = env.reset()
 
     steps = 0
     episodes = 0
diff --git a/nle/tests/test_envs.py b/nle/tests/test_envs.py
index 3e849a22b..1bf528af0 100644
--- a/nle/tests/test_envs.py
+++ b/nle/tests/test_envs.py
@@ -32,7 +32,7 @@ def rollout_env(env, max_rollout_len):
     Returns final reward. Does not assume that the environment has already been
     reset.
     """
-    obs = env.reset()
+    obs, reset_info = env.reset()
     assert env.observation_space.contains(obs)
 
     for _ in range(max_rollout_len):
@@ -97,14 +97,14 @@ def test_init(self, env_name, wizard):
     def test_reset(self, env_name, wizard):
         """Tests default initialization given standard env specs."""
         env = gym.make(env_name, wizard=wizard)
-        obs = env.reset()
+        obs, reset_info = env.reset()
         assert env.observation_space.contains(obs)
 
     def test_chars_colors_specials(self, env_name, wizard):
         env = gym.make(
             env_name, observation_keys=("chars", "colors", "specials", "blstats")
         )
-        obs = env.reset()
+        obs, reset_info = env.reset()
 
         assert "specials" in obs
         x, y = obs["blstats"][:2]
@@ -159,7 +159,7 @@ def test_meatball_exists(self):
         """Test loading stuff via wizkit"""
         env = gym.make("NetHack-v0", wizard=True)
         found = dict(meatball=0)
-        obs = env.reset(wizkit_items=list(found.keys()))
+        obs, reset_info = env.reset(wizkit_items=list(found.keys()))
         for line in obs["inv_strs"]:
             if np.all(line == 0):
                 break
@@ -199,7 +199,7 @@ def test_inventory(self, env_name):
                 "inv_oclasses",
             ),
         )
-        obs = env.reset()
+        obs, reset_info = env.reset()
 
         found = dict(spellbook=0, apple=0)
         for line in obs["inv_strs"]:
@@ -373,7 +373,7 @@ def test_kick_and_quit(self, env):
         assert reward == 0.0
 
     def test_final_reward(self, env):
-        obs = env.reset()
+        obs, reset_info = env.reset()
 
         for _ in range(100):
             obs, reward, done, info = env.step(env.action_space.sample())

From 64f9851a59458e2ea5c04bcc821d92752428647e Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Fri, 19 Apr 2024 17:30:02 +0100
Subject: [PATCH 12/17] Replace deprecated pkg_resources with
 importlib.resources

---
 nle/nethack/nethack.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/nle/nethack/nethack.py b/nle/nethack/nethack.py
index a9794847f..f6639aeec 100644
--- a/nle/nethack/nethack.py
+++ b/nle/nethack/nethack.py
@@ -1,4 +1,5 @@
 # Copyright (c) Facebook, Inc. and its affiliates.
+import importlib.resources
 import os
 import shutil
 import sys
@@ -7,7 +8,6 @@
 import weakref
 
 import numpy as np
-import pkg_resources
 
 from nle import _pynethack
 
@@ -67,7 +67,7 @@
     "time",
 )
 
-HACKDIR = pkg_resources.resource_filename("nle", "nethackdir")
+HACKDIR = importlib.resources.files("nle") / "nethackdir"
 TTYREC_VERSION = 3
 
 

From b79bf1cf4163eeb872fa3901a79f2ee2f7d26daa Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Sun, 21 Apr 2024 13:09:43 +0100
Subject: [PATCH 13/17] Move wizkit_items inside reset() 'options' parameter

---
 nle/env/base.py        |  8 +++-----
 nle/nethack/nethack.py | 13 +++++++------
 nle/tests/test_envs.py |  8 ++++----
 3 files changed, 14 insertions(+), 15 deletions(-)

diff --git a/nle/env/base.py b/nle/env/base.py
index d29d135b0..bdb00a41e 100644
--- a/nle/env/base.py
+++ b/nle/env/base.py
@@ -411,7 +411,7 @@ def _in_moveloop(self, observation):
         program_state = observation[self._program_state_index]
         return program_state[3]  # in_moveloop
 
-    def reset(self, seed=None, options=None, wizkit_items=None):
+    def reset(self, seed=None, options=None):
         """Resets the environment.
 
         Note:
@@ -431,9 +431,7 @@ def reset(self, seed=None, options=None, wizkit_items=None):
             new_ttyrec = self._ttyrec_pattern % self._episode
         else:
             new_ttyrec = None
-        self.last_observation = self.nethack.reset(
-            new_ttyrec, wizkit_items=wizkit_items
-        )
+        self.last_observation = self.nethack.reset(new_ttyrec, options=options)
 
         self._steps = 0
         done = False
@@ -453,7 +451,7 @@ def reset(self, seed=None, options=None, wizkit_items=None):
             warnings.warn(
                 "Not in moveloop after 1000 tries, aborting (ttyrec: %s)." % new_ttyrec
             )
-            return self.reset(wizkit_items=wizkit_items)
+            return self.reset(seed=seed, options=options)
 
         return self._get_observation(self.last_observation), self._get_information(
             self._get_end_status(self.last_observation, done)
diff --git a/nle/nethack/nethack.py b/nle/nethack/nethack.py
index f6639aeec..c801f2b24 100644
--- a/nle/nethack/nethack.py
+++ b/nle/nethack/nethack.py
@@ -250,12 +250,13 @@ def step(self, action):
         self._pynethack.step(action)
         return self._step_return(), self._pynethack.done()
 
-    def reset(self, new_ttyrec=None, wizkit_items=None):
-        if wizkit_items is not None:
-            if not self._wizard:
-                raise ValueError("Set wizard=True to use the wizkit option.")
-            # TODO ideally we need to check the validity of the requested items
-            self._pynethack.set_wizkit("\n".join(wizkit_items))
+    def reset(self, new_ttyrec=None, options=None):
+        if options is not None:
+            if options["wizkit_items"] is not None:
+                if not self._wizard:
+                    raise ValueError("Set wizard=True to use the wizkit option.")
+                # TODO ideally we need to check the validity of the requested items
+                self._pynethack.set_wizkit("\n".join(options["wizkit_items"]))
         if new_ttyrec is None:
             self._pynethack.reset()
         else:
diff --git a/nle/tests/test_envs.py b/nle/tests/test_envs.py
index 1bf528af0..d32ca369d 100644
--- a/nle/tests/test_envs.py
+++ b/nle/tests/test_envs.py
@@ -159,7 +159,7 @@ def test_meatball_exists(self):
         """Test loading stuff via wizkit"""
         env = gym.make("NetHack-v0", wizard=True)
         found = dict(meatball=0)
-        obs, reset_info = env.reset(wizkit_items=list(found.keys()))
+        obs, reset_info = env.reset(options={"wizkit_items": list(found.keys())})
         for line in obs["inv_strs"]:
             if np.all(line == 0):
                 break
@@ -173,16 +173,16 @@ def test_meatball_exists(self):
     def test_wizkit_no_wizard_mode(self):
         env = gym.make("NetHack-v0", wizard=False)
         with pytest.raises(ValueError) as e_info:
-            env.reset(wizkit_items=["meatball"])
+            env.reset(options={"wizkit_items": ["meatball"]})
         assert e_info.value.args[0] == "Set wizard=True to use the wizkit option."
 
     def test_wizkit_file(self):
         env = gym.make("NetHack-v0", wizard=True)
         req_items = ["meatball", "apple"]
-        env.reset(wizkit_items=req_items)
+        env.reset(options={"wizkit_items": req_items})
 
         # TODO: Test inventory here.
-        env.reset(wizkit_items=req_items)
+        env.reset(options={"wizkit_items": req_items})
         del env
 
 

From 4ded60c06608edb96bacf30ccca9274151c07495 Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Sun, 21 Apr 2024 18:54:41 +0100
Subject: [PATCH 14/17] Updates to support newer step() API

---
 nle/agent/agent.py        |  4 +++-
 nle/env/base.py           |  6 +++++-
 nle/scripts/play.py       |  2 +-
 nle/tests/test_envs.py    | 18 +++++++++---------
 nle/tests/test_profile.py |  2 +-
 nle/tests/test_system.py  |  6 +++---
 6 files changed, 22 insertions(+), 16 deletions(-)

diff --git a/nle/agent/agent.py b/nle/agent/agent.py
index 6498d11cf..1b14ebf11 100644
--- a/nle/agent/agent.py
+++ b/nle/agent/agent.py
@@ -362,7 +362,9 @@ def initial(self):
         return result
 
     def step(self, action):
-        observation, reward, done, unused_info = self.gym_env.step(action.item())
+        observation, reward, done, truncated, unused_info = self.gym_env.step(
+            action.item()
+        )
         self.episode_step += 1
         self.episode_return += reward
         episode_step = self.episode_step
diff --git a/nle/env/base.py b/nle/env/base.py
index bdb00a41e..0c02699d4 100644
--- a/nle/env/base.py
+++ b/nle/env/base.py
@@ -143,7 +143,7 @@ class NLE(gym.Env):
     Examples:
         >>> env = NLE()
         >>> obs, reset_info = env.reset()
-        >>> obs, reward, done, info = env.step(0)
+        >>> obs, reward, done, truncation, info = env.step(0)
         >>> env.render()
     """
 
@@ -371,6 +371,7 @@ def step(self, action: int):
                 - (*float*): a reward; see ``self._reward_fn`` to see how it is
                   specified.
                 - (*bool*): True if the state is terminal, False otherwise.
+                - (*bool*): True if the episode is truncated, False otherwise.
                 - (*dict*): a dictionary of extra information (such as
                   `end_status`, i.e. a status info -- death, task win, etc. --
                   for the terminal state).
@@ -400,10 +401,13 @@ def step(self, action: int):
             self._quit_game(observation, done)
             done = True
 
+        truncated = False
+
         return (
             self._get_observation(observation),
             reward,
             done,
+            truncated,
             self._get_information(end_status),
         )
 
diff --git a/nle/scripts/play.py b/nle/scripts/play.py
index 5d4b5f5d1..187992320 100644
--- a/nle/scripts/play.py
+++ b/nle/scripts/play.py
@@ -142,7 +142,7 @@ def play():
         if is_raw_env:
             obs, done = env.step(action)
         else:
-            obs, reward, done, info = env.step(action)
+            obs, reward, done, truncated, info = env.step(action)
         steps += 1
 
         if is_raw_env:
diff --git a/nle/tests/test_envs.py b/nle/tests/test_envs.py
index d32ca369d..c316b7a43 100644
--- a/nle/tests/test_envs.py
+++ b/nle/tests/test_envs.py
@@ -37,7 +37,7 @@ def rollout_env(env, max_rollout_len):
 
     for _ in range(max_rollout_len):
         a = env.action_space.sample()
-        obs, reward, done, info = env.step(a)
+        obs, reward, done, truncated, info = env.step(a)
         assert env.observation_space.contains(obs)
         assert isinstance(reward, float)
         assert isinstance(done, bool)
@@ -61,8 +61,8 @@ def compare_rollouts(env0, env1, max_rollout_len):
     step = 0
     while True:
         a = env0.action_space.sample()
-        obs0, reward0, done0, info0 = env0.step(a)
-        obs1, reward1, done1, info1 = env1.step(a)
+        obs0, reward0, done0, truncated0, info0 = env0.step(a)
+        obs1, reward1, done1, truncated1, info1 = env1.step(a)
         step += 1
 
         s0, s1 = term_screen(obs0), term_screen(obs1)
@@ -73,7 +73,7 @@ def compare_rollouts(env0, env1, max_rollout_len):
             np.testing.assert_equal(obs0, obs1)
         assert reward0 == reward1
         assert done0 == done1
-
+        assert truncated0 == truncated1
         assert info0 == info1
 
         if done0 or step >= max_rollout_len:
@@ -333,7 +333,7 @@ def test_render_ansi(self, env_name, rollout_len):
         env.reset()
         for _ in range(rollout_len):
             action = env.action_space.sample()
-            _, _, done, _ = env.step(action)
+            _, _, done, _, _ = env.step(action)
             if done:
                 env.reset()
             output = env.render()
@@ -361,13 +361,13 @@ def env(self):
     def test_kick_and_quit(self, env):
         env.reset()
         kick = env.actions.index(nethack.Command.KICK)
-        obs, reward, done, _ = env.step(kick)
+        obs, reward, done, _, _ = env.step(kick)
         assert b"In what direction? " in bytes(obs["message"])
         env.step(nethack.MiscAction.MORE)
 
         # Hack to quit.
         env.nethack.step(nethack.M("q"))
-        obs, reward, done, _ = env.step(env.actions.index(ord("y")))
+        obs, reward, done, _, _ = env.step(env.actions.index(ord("y")))
 
         assert done
         assert reward == 0.0
@@ -376,7 +376,7 @@ def test_final_reward(self, env):
         obs, reset_info = env.reset()
 
         for _ in range(100):
-            obs, reward, done, info = env.step(env.action_space.sample())
+            obs, reward, done, _, info = env.step(env.action_space.sample())
             if done:
                 break
 
@@ -391,7 +391,7 @@ def test_final_reward(self, env):
 
         # Hack to quit.
         env.nethack.step(nethack.M("q"))
-        _, reward, done, _ = env.step(env.actions.index(ord("y")))
+        _, reward, done, _, _ = env.step(env.actions.index(ord("y")))
 
         assert done
         assert reward == 0.0
diff --git a/nle/tests/test_profile.py b/nle/tests/test_profile.py
index 273d092bc..21be4bcca 100644
--- a/nle/tests/test_profile.py
+++ b/nle/tests/test_profile.py
@@ -64,7 +64,7 @@ def seed():
         def play_1k_steps():
             env.reset()
             for a in actions:
-                _, _, done, _ = env.step(a)
+                _, _, done, _, _ = env.step(a)
                 if done:
                     seed()
                     env.reset()
diff --git a/nle/tests/test_system.py b/nle/tests/test_system.py
index 0289c84be..375408a64 100644
--- a/nle/tests/test_system.py
+++ b/nle/tests/test_system.py
@@ -15,7 +15,7 @@
 def new_env_one_step():
     env = gym.make("NetHackScore-v0")
     env.reset()
-    obs, reward, done, _ = env.step(0)
+    obs, reward, done, _, _ = env.step(0)
     return done
 
 
@@ -50,7 +50,7 @@ def test_two_nles(self):
         num_resets = 1
 
         while num_resets < 4:
-            _, _, done, _ = env.step(random.choice(ACTIONS))
+            _, _, done, _, _ = env.step(random.choice(ACTIONS))
             if done:
                 queue.append(env)
                 env = queue.pop(0)
@@ -82,7 +82,7 @@ def target():
 
         while num_resets < 4:
             a = random.choice(ACTIONS)
-            _, _, done, _ = env.step(a)
+            _, _, done, _, _ = env.step(a)
             if done:
                 resetqueue.put(env)
                 env = readyqueue.get()

From 361db5a997565721d61d0b3ef364fca0b26cee84 Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Mon, 22 Apr 2024 16:10:31 +0100
Subject: [PATCH 15/17] Switch to Gymnasium

---
 .github/workflows/test_package.yml |  2 +-
 nle/agent/agent.py                 |  2 +-
 nle/env/__init__.py                |  4 +--
 nle/env/base.py                    |  2 +-
 nle/scripts/play.py                | 10 +++---
 nle/tests/test_db.py               |  2 +-
 nle/tests/test_envs.py             | 50 +++++++++++++++---------------
 nle/tests/test_profile.py          |  4 +--
 nle/tests/test_system.py           |  2 +-
 setup.py                           | 10 +++---
 10 files changed, 45 insertions(+), 43 deletions(-)

diff --git a/.github/workflows/test_package.yml b/.github/workflows/test_package.yml
index f501f7a42..fccefba57 100644
--- a/.github/workflows/test_package.yml
+++ b/.github/workflows/test_package.yml
@@ -27,4 +27,4 @@ jobs:
       run: "pip install nle"
     - name: Check nethack is installed
       run: |
-        python -c 'import nle, gym; e = gym.make("NetHack-v0"); e.reset(); e.step(0)' 
+        python -c 'import nle; import gymnasium as gym; e = gym.make("NetHack-v0"); e.reset(); e.step(0)' 
diff --git a/nle/agent/agent.py b/nle/agent/agent.py
index 1b14ebf11..f64ab7302 100644
--- a/nle/agent/agent.py
+++ b/nle/agent/agent.py
@@ -37,7 +37,7 @@
         '`pip install "nle[agent]"`'
     )
 
-import gym  # noqa: E402
+import gymnasium as gym  # noqa: E402
 
 import nle  # noqa: F401, E402
 from nle import nethack  # noqa: E402
diff --git a/nle/env/__init__.py b/nle/env/__init__.py
index 107c23198..03fa147a7 100644
--- a/nle/env/__init__.py
+++ b/nle/env/__init__.py
@@ -1,6 +1,6 @@
 # Copyright (c) Facebook, Inc. and its affiliates.
-import gym
-from gym.envs import registration
+import gymnasium as gym
+from gymnasium.envs import registration
 
 from nle.env.base import NLE, DUNGEON_SHAPE
 
diff --git a/nle/env/base.py b/nle/env/base.py
index 0c02699d4..68bc665fa 100644
--- a/nle/env/base.py
+++ b/nle/env/base.py
@@ -9,7 +9,7 @@
 import warnings
 import weakref
 
-import gym
+import gymnasium as gym
 import numpy as np
 
 from nle import nethack
diff --git a/nle/scripts/play.py b/nle/scripts/play.py
index 187992320..d4fd6c938 100644
--- a/nle/scripts/play.py
+++ b/nle/scripts/play.py
@@ -11,7 +11,7 @@
 import timeit
 import tty
 
-import gym
+import gymnasium as gym
 
 import nle  # noqa: F401
 from nle import nethack
@@ -60,7 +60,7 @@ def get_action(env, is_raw_env):
                 if is_raw_env:
                     action = ch
                 else:
-                    action = env.actions.index(ch)
+                    action = env.unwrapped.actions.index(ch)
                 break
             except ValueError:
                 print(
@@ -95,7 +95,7 @@ def play():
             render_mode=FLAGS.render_mode,
         )
         if FLAGS.seeds is not None:
-            env.seed(FLAGS.seeds)
+            env.unwrapped.seed(FLAGS.seeds)
 
     obs, reset_info = env.reset()
 
@@ -115,7 +115,9 @@ def play():
             if not is_raw_env:
                 print("-" * 8 + " " * 71)
                 print(f"Previous reward: {str(reward):64s}")
-                act_str = repr(env.actions[action]) if action is not None else ""
+                act_str = (
+                    repr(env.unwrapped.actions[action]) if action is not None else ""
+                )
                 print(f"Previous action: {str(act_str):64s}")
                 print("-" * 8)
                 env.render()
diff --git a/nle/tests/test_db.py b/nle/tests/test_db.py
index e370cb0ed..770cada33 100644
--- a/nle/tests/test_db.py
+++ b/nle/tests/test_db.py
@@ -26,7 +26,7 @@ def gen_ttyrecs(savedir):
         for _ in range(5):
             # Need to end naturally to be recorded.
             for c in [ord(" "), ord(" "), ord("<"), ord("y")]:
-                _, _, done, *_ = env.step(env.actions.index(c))
+                _, _, done, *_ = env.step(env.unwrapped.actions.index(c))
             assert done
             env.reset()
         env.close()
diff --git a/nle/tests/test_envs.py b/nle/tests/test_envs.py
index c316b7a43..f356038de 100644
--- a/nle/tests/test_envs.py
+++ b/nle/tests/test_envs.py
@@ -7,7 +7,7 @@
 import sys
 import tempfile
 
-import gym
+import gymnasium as gym
 import numpy as np
 import pytest
 
@@ -120,11 +120,11 @@ def test_default_wizard_mode(self, env_name, wizard):
             if env_name.startswith("NetHackChallenge-"):
                 pytest.skip("No wizard mode in NetHackChallenge")
             env = gym.make(env_name, wizard=wizard)
-            assert "playmode:debug" in env.nethack.options
+            assert "playmode:debug" in env.unwrapped.nethack.options
         else:
             # do not send a parameter to test a default
             env = gym.make(env_name)
-            assert "playmode:debug" not in env.nethack.options
+            assert "playmode:debug" not in env.unwrapped.nethack.options
 
 
 class TestWizardMode:
@@ -254,7 +254,7 @@ def test_rollout(self, env_name, rollout_len):
     def test_rollout_no_archive(self, env_name, rollout_len):
         """Tests rollout_len steps (or until termination) of random policy."""
         env = gym.make(env_name, savedir=None)
-        assert env.savedir is None
+        assert env.unwrapped.savedir is None
         rollout_env(env, rollout_len)
 
     def test_seed_interface_output(self, env_name, rollout_len):
@@ -267,12 +267,12 @@ def test_seed_interface_output(self, env_name, rollout_len):
         env0 = gym.make(env_name)
         env1 = gym.make(env_name)
 
-        seed_list0 = env0.seed()
+        seed_list0 = env0.unwrapped.seed()
         env0.reset()
 
-        assert env0.get_seeds() == seed_list0
+        assert env0.unwrapped.get_seeds() == seed_list0
 
-        seed_list1 = env1.seed(*seed_list0)
+        seed_list1 = env1.unwrapped.seed(*seed_list0)
         assert seed_list0 == seed_list1
 
     def test_seed_rollout_seeded(self, env_name, rollout_len):
@@ -285,15 +285,15 @@ def test_seed_rollout_seeded(self, env_name, rollout_len):
         env0 = gym.make(env_name)
         env1 = gym.make(env_name)
 
-        env0.seed(123456, 789012)
+        env0.unwrapped.seed(123456, 789012)
         obs0 = env0.reset()
-        seeds0 = env0.get_seeds()
+        seeds0 = env0.unwrapped.get_seeds()
 
         assert seeds0 == (123456, 789012, False)
 
-        env1.seed(*seeds0)
+        env1.unwrapped.seed(*seeds0)
         obs1 = env1.reset()
-        seeds1 = env1.get_seeds()
+        seeds1 = env1.unwrapped.get_seeds()
 
         assert seeds0 == seeds1
 
@@ -315,13 +315,13 @@ def test_seed_rollout_seeded_int(self, env_name, rollout_len):
             random.randrange(sys.maxsize),
             False,
         )
-        env0.seed(*initial_seeds)
+        env0.unwrapped.seed(*initial_seeds)
         obs0 = env0.reset()
-        seeds0 = env0.get_seeds()
+        seeds0 = env0.unwrapped.get_seeds()
 
-        env1.seed(*seeds0)
+        env1.unwrapped.seed(*seeds0)
         obs1 = env1.reset()
-        seeds1 = env1.get_seeds()
+        seeds1 = env1.unwrapped.get_seeds()
 
         assert seeds0 == seeds1 == initial_seeds
 
@@ -360,14 +360,14 @@ def env(self):
 
     def test_kick_and_quit(self, env):
         env.reset()
-        kick = env.actions.index(nethack.Command.KICK)
+        kick = env.unwrapped.actions.index(nethack.Command.KICK)
         obs, reward, done, _, _ = env.step(kick)
         assert b"In what direction? " in bytes(obs["message"])
         env.step(nethack.MiscAction.MORE)
 
         # Hack to quit.
-        env.nethack.step(nethack.M("q"))
-        obs, reward, done, _, _ = env.step(env.actions.index(ord("y")))
+        env.unwrapped.nethack.step(nethack.M("q"))
+        obs, reward, done, _, _ = env.step(env.unwrapped.actions.index(ord("y")))
 
         assert done
         assert reward == 0.0
@@ -387,11 +387,11 @@ def test_final_reward(self, env):
         # Hopefully, we got some positive reward by now.
 
         # Get out of any menu / yn_function.
-        env.step(env.actions.index(ord("\r")))
+        env.step(env.unwrapped.actions.index(ord("\r")))
 
         # Hack to quit.
-        env.nethack.step(nethack.M("q"))
-        _, reward, done, _, _ = env.step(env.actions.index(ord("y")))
+        env.unwrapped.nethack.step(nethack.M("q"))
+        _, reward, done, _, _ = env.step(env.unwrapped.actions.index(ord("y")))
 
         assert done
         assert reward == 0.0
@@ -403,7 +403,7 @@ def test_ttyrec_every(self):
         for episode in range(10):
             env.reset()
             for c in [ord(" "), ord(" "), ord("<"), ord("y")]:
-                _, _, done, *_ = env.step(env.actions.index(c))
+                _, _, done, *_ = env.step(env.unwrapped.actions.index(c))
             assert done
 
             if episode % 2 != 0:
@@ -449,10 +449,10 @@ def test_no_seed_setting(self):
         with pytest.raises(
             RuntimeError, match="NetHackChallenge doesn't allow seed changes"
         ):
-            env.seed()
+            env.unwrapped.seed()
         with pytest.raises(RuntimeError, match="Should not try changing seeds"):
-            env.nethack.set_initial_seeds(0, 0, True)
+            env.unwrapped.nethack.set_initial_seeds(0, 0, True)
 
         if not nethack.NLE_ALLOW_SEEDING:
             with pytest.raises(RuntimeError, match="Seeding not enabled"):
-                env.nethack._pynethack.set_initial_seeds(0, 0, True)
+                env.unwrapped.nethack._pynethack.set_initial_seeds(0, 0, True)
diff --git a/nle/tests/test_profile.py b/nle/tests/test_profile.py
index 21be4bcca..220f385bb 100644
--- a/nle/tests/test_profile.py
+++ b/nle/tests/test_profile.py
@@ -7,7 +7,7 @@
 # to run
 import os
 
-import gym
+import gymnasium as gym
 import numpy as np
 import pytest
 
@@ -59,7 +59,7 @@ def seed():
                 return
             nonlocal seeds
             seeds += 1
-            env.seed(seeds, 2 * seeds)
+            env.unwrapped.seed(seeds, 2 * seeds)
 
         def play_1k_steps():
             env.reset()
diff --git a/nle/tests/test_system.py b/nle/tests/test_system.py
index 375408a64..b7c79df67 100644
--- a/nle/tests/test_system.py
+++ b/nle/tests/test_system.py
@@ -4,7 +4,7 @@
 import random
 import threading
 
-import gym
+import gymnasium as gym
 import pytest
 
 import nle  # noqa: F401
diff --git a/setup.py b/setup.py
index f48360a86..979bf0ede 100644
--- a/setup.py
+++ b/setup.py
@@ -163,19 +163,19 @@ def build_extension(self, ext):
         ext_modules=[setuptools.Extension("nle", sources=[])],
         cmdclass={"build_ext": CMakeBuild},
         setup_requires=["pybind11>=2.2"],
-        install_requires=["pybind11>=2.2", "numpy>=1.16", "gym>=0.15"],
+        install_requires=["pybind11>=2.2", "numpy>=1.16", "gymnasium>=0.29.1"],
         extras_require=extras_deps,
-        python_requires=">=3.5",
+        python_requires=">=3.8",
         classifiers=[
             "License :: OSI Approved :: Nethack General Public License",
             "Development Status :: 4 - Beta",
             "Operating System :: POSIX :: Linux",
             "Operating System :: MacOS",
             "Programming Language :: Python :: 3",
-            "Programming Language :: Python :: 3.5",
-            "Programming Language :: Python :: 3.6",
-            "Programming Language :: Python :: 3.7",
             "Programming Language :: Python :: 3.8",
+            "Programming Language :: Python :: 3.9",
+            "Programming Language :: Python :: 3.10",
+            "Programming Language :: Python :: 3.11",
             "Programming Language :: C",
             "Programming Language :: C++",
             "Topic :: Scientific/Engineering :: Artificial Intelligence",

From 818d24909773523d37fb54d892fa69cd8055d505 Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Tue, 23 Apr 2024 09:13:49 +0100
Subject: [PATCH 16/17] Fix to enable importlib.resources in Python 3.8

---
 nle/nethack/nethack.py | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/nle/nethack/nethack.py b/nle/nethack/nethack.py
index c801f2b24..d58188f4e 100644
--- a/nle/nethack/nethack.py
+++ b/nle/nethack/nethack.py
@@ -67,7 +67,16 @@
     "time",
 )
 
-HACKDIR = importlib.resources.files("nle") / "nethackdir"
+# "The past is a foreign country, they do things differently there."
+#       - L. P. Hartley (1895 - 1972), "The Go-Between"
+#
+# importlib.resources API was indeed different in Python 3.8.
+if sys.version_info < (3, 9):
+    with importlib.resources.path("nle", "nethackdir") as nh:
+        HACKDIR = os.path.abspath(nh)
+else:
+    HACKDIR = importlib.resources.files("nle") / "nethackdir"
+
 TTYREC_VERSION = 3
 
 

From 8cb2d13e701b1deb35bf47a12493717329223c7c Mon Sep 17 00:00:00 2001
From: Stephen Oman <stephen.oman@yahoo.co.uk>
Date: Tue, 23 Apr 2024 19:30:30 +0100
Subject: [PATCH 17/17] Update README.md with Gymnasium instructions

---
 README.md | 13 +++++++------
 1 file changed, 7 insertions(+), 6 deletions(-)

diff --git a/README.md b/README.md
index bb18d9ff2..b465f8d74 100644
--- a/README.md
+++ b/README.md
@@ -76,12 +76,12 @@ to add papers.
 # Getting started
 
 Starting with NLE environments is extremely simple, provided one is familiar
-with other gym / RL environments.
+with other Gymnasium / RL environments.
 
 
 ## Installation
 
-NLE requires `python>=3.5`, `cmake>=3.15` to be installed and available both when building the
+NLE requires `python>=3.8`, `cmake>=3.15` to be installed and available both when building the
 package, and at runtime.
 
 On **MacOS**, one can use `Homebrew` as follows:
@@ -136,7 +136,7 @@ README](docker/README.md).
 After installation, one can try out any of the provided tasks as follows:
 
 ```python
->>> import gym
+>>> import gymnasium as gym
 >>> import nle
 >>> env = gym.make("NetHackScore-v0")
 >>> env.reset()  # each reset generates a new dungeon
@@ -174,8 +174,9 @@ $ pip install "nle[agent]"
 $ python -m nle.agent.agent --num_actors 80 --batch_size 32 --unroll_length 80 --learning_rate 0.0001 --entropy_cost 0.0001 --use_lstm --total_steps 1000000000
 ```
 
-Plot the mean return over the last 100 episodes:
+Plot the mean return over the last 100 episodes (requires gnuplotlib):
 ```bash
+$ pip install gnuplotlib
 $ python -m nle.scripts.plot
 ```
 ```
@@ -222,8 +223,8 @@ see [this document](./CONTRIBUTING.md).
 NLE is direct fork of [NetHack](https://github.com/nethack/nethack) and
 therefore contains code that operates on many different levels of abstraction.
 This ranges from low-level game logic, to the higher-level administration of
-repeated nethack games, and finally to binding of these games to Python `gym`
-environment.
+repeated nethack games, and finally to binding of these games to Python 
+`gymbasium` environment.
 
 If you want to learn more about the architecture of `nle` and how it works
 under the hood, checkout the [architecture document](./doc/nle/ARCHITECTURE.md).