fix(pypi): stricter requirements for dependencies (#76)

- update mathy_core and mathy_envs to latest version that pin gym to the required range

Author: justindujardin

libraries/mathy_python/mathy/api.py

@@ -1,4 +1,5 @@
 from dataclasses import dataclass
+from typing import Optional
 
 from fragile.core.swarm import Swarm
 
@@ -16,15 +17,18 @@ class Mathy:
     state: MathyAPISwarmState
 
     def __init__(
-        self, *, config: SwarmConfig = None, silent: bool = False,
+        self,
+        *,
+        config: Optional[SwarmConfig] = None,
+        silent: bool = False,
     ):
         if config is None:
             config = SwarmConfig()
         if not isinstance(config, SwarmConfig):
             raise ValueError("config must be a SwarmConfig instance")
         self.state = MathyAPISwarmState(config=config)
 
-    def simplify(self, *, problem: str, max_steps: int = None) -> Swarm:
+    def simplify(self, *, problem: str, max_steps: Optional[int] = None) -> Swarm:
         if max_steps is not None:
             return swarm_solve(problem, self.state.config, max_steps=max_steps)
         return swarm_solve(problem, self.state.config)

libraries/mathy_python/mathy/solver.py

@@ -1,6 +1,6 @@
 """Use Fractal Monte Carlo search in order to solve mathy problems without a
 trained neural network."""
-from typing import Dict, List, Optional, Union
+from typing import Any, Dict, List, Optional, Union
 
 import numpy as np
 from fragile.core.env import DiscreteEnv
@@ -70,7 +70,7 @@ def __init__(
         name: str,
         environment: str = "poly",
         difficulty: str = "easy",
-        problem: str = None,
+        problem: Optional[str] = None,
         max_steps: int = 64,
         **kwargs,
     ):
@@ -122,7 +122,7 @@ def __init__(
         name: str,
         environment: str = "poly",
         difficulty: str = "normal",
-        problem: str = None,
+        problem: Optional[str] = None,
         max_steps: int = 64,
         **kwargs,
     ):
@@ -132,10 +132,9 @@ def __init__(
 
         self._env: MathyGymEnv = gym.make(
             f"mathy-{environment}-{difficulty}-v0",
-            np_observation=True,
-            mask_as_probabilities=True,
             invalid_action_response="terminal",
             env_problem=problem,
+            mask_as_probabilities=True,
             **kwargs,
         )
         self.observation_space = spaces.Box(
@@ -148,7 +147,7 @@ def __init__(
 
     def get_state(self) -> np.ndarray:
         assert self._env.state is not None, "env required to get_state"
-        return self._env.state.to_np()
+        return self._env.state.to_np(768)
 
     def set_state(self, state: np.ndarray):
         assert self._env is not None, "env required to set_state"
@@ -165,7 +164,7 @@ def step(self, action: int, state: np.ndarray = None) -> tuple:
         return new_state, obs, reward, oob, info
 
     def step_batch(
-        self, actions, states=None, n_repeat_action: Union[int, np.ndarray] = None
+        self, actions, states:Optional[Any]=None, n_repeat_action: Optional[Union[int, np.ndarray]] = None
     ) -> tuple:
         data = [self.step(action, state) for action, state in zip(actions, states)]
         new_states, observs, rewards, terminals, infos = [], [], [], [], []
@@ -212,7 +211,7 @@ def mathy_swarm(config: SwarmConfig, env_callable=None) -> Swarm:
 def swarm_solve(
     problems: Union[List[str], str],
     config: SwarmConfig,
-    max_steps: Union[List[int], int] = 128,
+    max_steps: Union[List[int], int] = 256,
     silent: bool = False,
 ) -> Swarm:
     single_problem: bool = isinstance(problems, str)

libraries/mathy_python/requirements-dev.txt

@@ -1,7 +1,7 @@
 mypy
 pytest
 pytest-cov
-black
+black<22.2.0
 isort
 autoflake
 flake8

libraries/mathy_python/requirements.txt

@@ -1,12 +1,12 @@
 # Core library
-numpy<1.19.0,>=1.16.0
+numpy<1.23.0,>=1.22.0
 colr
 srsly
 pydantic>=1.0.0
-typer<0.4.0,>=0.3.2
+typer<0.7.0,>=0.6.1
 wasabi
-mathy_core>=0.8.4,<0.9.0
-mathy_envs[gym]>=0.10.0,<0.11.0
+mathy_core>=0.8.6,<0.9.0
+mathy_envs[gym]>=0.11.3,<0.12.0
 fragile==0.0.47
 tqdm>=4.43.0
 # new python feature backports