Open RasheedOlaleye opened 2 weeks ago
Sadly, whenever specifying "CUDA Toolkit" one must specify "cuda-toolkit". This is because there is something called "cudatoolkit" that silently poison's one's environment and people often confuse the two.
(xLSTM) jabowery@jaboweryML:~/devel/xLSTM$ ./build.sh -- Configuring done CMake Error in CMakeLists.txt: CUDA_ARCHITECTURES is empty for target "xlstm_cuda".
CMake Error in CMakeLists.txt: CUDA_ARCHITECTURES is empty for target "xlstm_cuda".
Addressing CUDA_ARCHITECTURES problem by altering build.sh
to contain this cmake
call:
(xLSTM) jabowery@jaboweryML:~/devel/xLSTM/cuda/build$ cmake .. -DUSE_CUDA=ON -DCMAKE_CUDA_ARCHITECTURES=86 -DCMAKE_BUILD_TYPE=Release
-- Configuring done
-- Generating done
CMake Warning:
Manually-specified variables were not used by the project:
USE_CUDA
-- Build files have been written to: /home/jabowery/devel/xLSTM/cuda/build
(xLSTM) jabowery@jaboweryML:~/devel/xLSTM/cuda/build$ ls
CMakeCache.txt CMakeFiles cmake_install.cmake Makefile
(xLSTM) jabowery@jaboweryML:~/devel/xLSTM/cuda/build$ make
[ 25%] Building CUDA object CMakeFiles/xlstm_cuda.dir/kernels/slstm_kernels.cu.o
/home/jabowery/devel/xLSTM/cuda/kernels/slstm_kernels.cu(111): error: identifier "b_i" is undefined
b_i[hidx]);
^
/home/jabowery/devel/xLSTM/cuda/kernels/slstm_kernels.cu(114): error: identifier "b_f" is undefined
b_f[hidx]);
^
/home/jabowery/devel/xLSTM/cuda/kernels/slstm_kernels.cu(117): error: identifier "b_z" is undefined
b_z[hidx]);
^
/home/jabowery/devel/xLSTM/cuda/kernels/slstm_kernels.cu(120): error: identifier "b_o" is undefined
b_o[hidx]);
^
4 errors detected in the compilation of "/home/jabowery/devel/xLSTM/cuda/kernels/slstm_kernels.cu".
make[2]: *** [CMakeFiles/xlstm_cuda.dir/build.make:77: CMakeFiles/xlstm_cuda.dir/kernels/slstm_kernels.cu.o] Error 2
make[1]: *** [CMakeFiles/Makefile2:83: CMakeFiles/xlstm_cuda.dir/all] Error 2
make: *** [Makefile:91: all] Error 2
To help you set up the xLSTM project, I'll provide a step-by-step guide based on the information available in the codebase. This guide will cover the installation and setup process for both the C++ and Python components of the project.
Step 1: Prerequisites
Ensure you have the following prerequisites installed on your system:
Step 2: Clone the Repository
If you haven't already, clone the xLSTM repository to your local machine:
Step 3: Build the C++ Components
The project includes a shell script
build.sh
that automates the build process for the C++ components. Run the following command to build and install the xLSTM library:This script will:
Step 4: Install Python Dependencies
Navigate to the
python
directory and install the required Python dependencies usingpip
:Step 5: Build and Install the Python Package
Still in the
python
directory, build and install the xLSTM Python package:Step 6: Verify the Installation
To verify that the installation was successful, you can run one of the provided Python examples. For instance, to run the language model example:
This script will:
Additional Information
For more detailed information on the xLSTM architecture, usage examples, and contributing guidelines, refer to the
README.md
andOVERVIEW.md
files.Troubleshooting
If you encounter any issues during the setup process, consider the following:
README.md
file for any additional setup instructions or requirements.By following these steps, you should be able to set up the xLSTM project successfully. If you have any further questions or run into specific issues, feel free to ask!