Can not use docker

[Haoran-P]

Thanks for the great work!
I am using HEC platform ,whichdoes not support container such as docker.
May I ask if there is other method to run the code without container.

[hung61601]

Hi, thanks for your interest in this project!
Although the program currently cannot run out-of-the-box without a container, you can still execute it on your system by manually installing the required dependencies.
To help with this, we've provided the Dockerfile (https://github.com/rlglab/minizero/blob/main/docker/Dockerfile) as a reference.
Please refer to it to install the necessary packages on your system (e.g., CUDA, Python packages, and other tools).
The Dockerfile was added in commit 2974f58 as part of the latest update.

[PercyLau]

Quite weird in the dockerfile:

RUN python3 -m pip install torch==1.13.0+cu116 --extra-index-url https://download.pytorch.org/whl/cu116 &&
wget -qO- https://download.pytorch.org/libtorch/cu116/libtorch-cxx11-abi-shared-with-deps-1.13.0%2Bcu116.zip | bsdtar -xvf- -C /usr/local &&
rm /usr/local/libtorch/lib/libcudnn.so.8
ENV CMAKE_PREFIX_PATH /usr/local/libtorch/share/cmake/Torch.

why should we delete the so？

[PercyLau]

/workspace# tools/quick-run.sh train go az 300 -n go_9x9_az_n200 -conf_str env_board_size=9:actor_num_simulation=200
MiniZero quick-run session: minizero_go.2974f5-dirty.train.20250720153548
Already running inside the container, will launch the program directly
Building the executable for go ...
game type: go
build type: Release
-- The C compiler identification is GNU 9.4.0
-- The CXX compiler identification is GNU 9.4.0
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: /usr/bin/cc - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/c++ - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Found Python: /usr/bin/python3.8 (found version "3.8.10") found components: Interpreter Development Development.Module Development.Embed
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Failed
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - found
-- Found Threads: TRUE
-- Found CUDA: /usr/local/cuda (found version "11.6")
-- The CUDA compiler identification is NVIDIA 11.6.124 with host compiler GNU 9.4.0
-- Detecting CUDA compiler ABI info
-- Detecting CUDA compiler ABI info - done
-- Check for working CUDA compiler: /usr/local/cuda/bin/nvcc - skipped
-- Detecting CUDA compile features
-- Detecting CUDA compile features - done
-- Caffe2: CUDA detected: 11.6
-- Caffe2: CUDA nvcc is: /usr/local/cuda/bin/nvcc
-- Caffe2: CUDA toolkit directory: /usr/local/cuda
-- Caffe2: Header version is: 11.6
-- Found CUDNN: /usr/lib/x86_64-linux-gnu/libcudnn.so
-- Found cuDNN: v8.4.0 (include: /usr/include, library: /usr/lib/x86_64-linux-gnu/libcudnn.so)
-- /usr/local/cuda/lib64/libnvrtc.so shorthash is 4dd39364
-- Autodetected CUDA architecture(s): 8.9 8.9
-- Added CUDA NVCC flags for: -gencode;arch=compute_89,code=sm_89
-- Found Torch: /usr/local/libtorch/lib/libtorch.so
CMake Warning (dev) at CMakeLists.txt:11 (find_package):
Policy CMP0167 is not set: The FindBoost module is removed. Run "cmake
--help-policy CMP0167" for policy details. Use the cmake_policy command to
set the policy and suppress this warning.

This warning is for project developers. Use -Wno-dev to suppress it.

-- Found Boost: /usr/lib/x86_64-linux-gnu/cmake/Boost-1.71.0/BoostConfig.cmake (found version "1.71.0") found components: system thread iostreams
-- Found ZLIB: /usr/lib/x86_64-linux-gnu/libz.so (found version "1.2.11")
-- Found OpenCV: /usr (found version "4.2.0")
CMake Error at /usr/lib/cmake/pybind11/pybind11Tools.cmake:8 (cmake_minimum_required):
Compatibility with CMake < 3.5 has been removed from CMake.

Update the VERSION argument value. Or, use the ... syntax
to tell CMake that the project requires at least but has been updated
to work with policies introduced by or earlier.

Or, add -DCMAKE_POLICY_VERSION_MINIMUM=3.5 to try configuring anyway.
Call Stack (most recent call first):
/usr/lib/cmake/pybind11/pybind11Config.cmake:100 (include)
minizero/learner/CMakeLists.txt:3 (find_package)

-- Configuring incomplete, errors occurred!
Failed to build the executable for go

---

this dockerfile will generate the above error

[moporgic]

The issue seems to be related to the CMake version. Could you help verify which version of CMake you are currently using?
Our container image uses CMake 3.25.2, with the CMakeLists.txt specifying cmake_minimum_required(VERSION 3.16).
However, the log Compatibility with CMake < 3.5 has been removed from CMake. indicates that you are likely using a higher version of CMake, causing the error to be raised.
If so, please try changing cmake_minimum_required(VERSION 3.16) to cmake_minimum_required(VERSION 3.5) and check again. Thanks!

[PercyLau]

I think you are right. The dockerfile somehow installs two versions of cmake in my case. The first one is from apt-get and the later by pip install, as you can check,

python3 -m pip install -U pip \ gdown \ cmake \

The above script will lead to CMake 4.0.rc. @moporgic

[moporgic]

@hung61601, could you help resolve this issue? We might need to explicitly specify the versions for the required packages, or remove the duplicate CMake installation.

[hung61601]

@PercyLau Thanks for pointing that out!
The line rm /usr/local/libtorch/lib/libcudnn.so.8 is meant to avoid potential conflicts between the bundled cuDNN and the system’s existing CUDA libraries.
In some environments, using the prepackaged libcudnn.so.8 can lead to issues like version mismatches or runtime errors, so we remove it to ensure compatibility with the system-installed CUDA setup.
As for CMake, installing a newer version via pip is necessary because the one from apt-get doesn’t support C++17.
However, CMake 4.0 (installed via pip) is also problematic—it doesn't work well with older Pybind11 versions.
We resolved this in the latest commit 04a5895, so if you rebuild using the updated Dockerfile, it should work correctly now.