bidexhands_validation

Docker Image for Bi-DexHands Domain

Installation Phases

1. Installing PyTorch and flash_attn
2. Patching IsaacGym torch_utils.py file in order to be compatible with numpy>=1.24.4
3. Installing IsaacGym
4. Patching and installing Bi-DexHands

Initializing Bi-DexHands Gym environment

In newly created environment run these commands:

from envs.wrappers.bidexhands import get_gym_env, TASKS

print(f"Available tasks are:", *TASKS, sep='\n')

# Initialize env
env = get_gym_env(task_name="ShadowHandCatchUnderarm")

# Don`t forget to .close before initializing new env!
env.close()

Notes

This Docker image supports only CPU-based environments due to the lack of Vulcan installation.