Tri Wahyu Guntara

Hi @ynulihao , may I know the code you are running to get `stable-baselines3` baseline?

Any updates? Anyone successfully reproduce the results from the paper?

Perhaps we can use [`shapely`](https://shapely.readthedocs.io/en/stable/manual.html) to compute intersection area between two polygons. Quick example is [here](https://stackoverflow.com/a/57885693). Details: Changing ``` intersection_points = polygon_intersection(a.p[0:4, 0:2], b.p[0:4, 0:2]) inter_area = polygon_area(intersection_points) ``` to...

I think it works as intended. `metaworld` has two types of variations, parametric variations (goal, initial hand position, initial object position) and non-parametric variations (eg. hammer and reach). I will...

I have similar issue. I only tested on the MT10 tasks and train a policy on each of the tasks. I run evaluation every 160k steps in the environment where...

To reproduce, here is the code ```Python3 from typing import Any, Dict, List, Tuple import gym import metaworld import numpy as np import os.path import random from gym.wrappers import TimeLimit...

I found the culprit. 1. The `metaworld.Benchmark` object should only be created once. Recreating the benchmark object will sample a new tasks (ie. start/object/goal positions). In my code, I call...

@UditSinghParihar you also have to change the default config to spawn ego if you use that one ([here](https://github.com/carla-simulator/ros-bridge/blob/master/carla_spawn_objects/config/objects.json#L40)). You have to negate the pitch. This is very very inconvenient.