Install Isaac Gym on ubuntu 22.04

isaac_gym_ubuntu22.04.md

1. Install Miniconda

mkdir -p ~/miniconda3
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
rm -rf ~/miniconda3/miniconda.sh
~/miniconda3/bin/conda init bash

2. Install Isaac Gym

Download Isaac Gym

Unzip and install

cd ~/Downloads
mv IsaacGym_Preview_4_Package.tar.gz ~/.
cd ~
tar -xf IsaacGym_Preview_4_Package.tar.gz
cd isaacgym
./create_conda_env_rlgpu.sh

Once it finishes, it will show SUCCESS.

3. Test installation

Activate conda env

conda activate rlgpu

Run an example

cd ~/isaacgym/python/examples
python joint_monkey.py

If it doesn't run and complain about some plugins, it is likely that you need to install an Nvidia driver for your GPU. See NVIDIA-Ubuntu-Driver-Guide.

4. Install the IsaacGymEnvs

git clone https://github.com/NVIDIA-Omniverse/IsaacGymEnvs.git
cd IsaacGymEnvs
pip install -e .

5. Run a training example

cd ~/IsaacGymEnvs/isaacgymenvs
python3 train.py task=Cartpole

We can also set number of environment with the num_envs argument

python train.py task=BallBalance num_envs=512 train.params.config.minibatch_size=512

6. Troubleshooting

libpython3.7

Error message:

ImportError: libpython3.7m.so.1.0: cannot open shared object file: No such file or directory

Solution:

sudo apt install libpython3.7
export LD_LIBRARY_PATH=~/miniconda3/envs/rlgpu/lib:$LD_LIBRARY_PATH

libcublasLt.so.11

Error message:

OSError: /home/anthony/miniconda3/envs/rlgpu/lib/python3.7/site-packages/nvidia/cublas/lib/libcublas.so.11: undefined symbol: cublasLtHSHMatmulAlgoInit, version libcublasLt.so.11

Solution:

sudo apt install libpython3.7
export LD_LIBRARY_PATH=~/miniconda3/envs/rlgpu/lib/python3.7/site-packages/nvidia/cublas/lib/:$LD_LIBRARY_PATH

undefined symbol: iJIT_NotifyEvent

Error message:

ImportError: /home/anthony/miniconda3/envs/rlgpu/lib/python3.7/site-packages/torch/lib/libtorch_cpu.so: undefined symbol: iJIT_NotifyEvent

Solution:

pip install mkl==2024.0.0

undefined symbol: cublasLtHSHMatmulAlgoInit

Error message:

undefined symbol: cublasLtHSHMatmulAlgoInit, version libcublasLt.so.11

This happens because pytorch1.13 automatically installed nvidia_cublas_cu11, nvidia_cuda_nvrtc_cu11, nvidia_cuda_runtime_cu11 and nvidia_cudnn_cu11.

Solution:

pip uninstall nvidia_cublas_cu11

Credit

This is a cleaned up copy of William Chen's work