hongshaorou/tdmpc2
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

cleanup

Browse Source
This commit is contained in:
Nicklas Hansen
2024-11-11 19:09:09 -08:00
parent dc6720d322
commit a9b5ad0ff8
1 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
tdmpc2/envs/wrappers/discrete.py
View File

@@ -1,5 +1,4 @@
import gym import gym
import numpy as np
import torch import torch
from common import math from common import math
@@ -10,12 +9,13 @@ class DiscreteWrapper(gym.Wrapper):
Wrapper for converting continuous action spaces to discrete via binning. Wrapper for converting continuous action spaces to discrete via binning.
""" """
def __init__(self, env): def __init__(self, env, bins_per_dim=5):
super().__init__(env) super().__init__(env)
self.bins_per_dim = bins_per_dim
self.continuous_dims = self.env.action_space.shape[0] self.continuous_dims = self.env.action_space.shape[0]
# Bins at [-1, 0, 1] for each dimension # Bins at [-1, 0, 1] for each dimension
# Discrete actions include all possible combinations of these bins # Discrete actions include all possible combinations of these bins
self.action_space = gym.spaces.Discrete(3 ** self.continuous_dims) self.action_space = gym.spaces.Discrete(bins_per_dim ** self.continuous_dims)
def rand_act(self): def rand_act(self):
action = torch.tensor(self.action_space.sample(), dtype=torch.int64) action = torch.tensor(self.action_space.sample(), dtype=torch.int64)
@@ -26,7 +26,7 @@ class DiscreteWrapper(gym.Wrapper):
# action is a one-hot encoded tensor # action is a one-hot encoded tensor
action = torch.argmax(action) action = torch.argmax(action)
action = action.item() action = action.item()
action = [action // 3 ** i % 3 for i in range(self.continuous_dims)] action = [action // self.bins_per_dim ** i % self.bins_per_dim for i in range(self.continuous_dims)]
action = torch.tensor(action, dtype=torch.float32) action = torch.tensor(action, dtype=torch.float32)
return (action - 1) / 1 return (action - 1) / 1