ERROR: No matching distribution found for BAcore

[stiiff]

There is no BACore module in pytorch？

[nesi73]

The initial issue was that I was compiling with the wrong version of GCC, version 13, while the correct one was version 10. To ensure this, I verified that gcc --version and g++ --version both returned version 10. Once I fixed this, the BACore error was resolved, but I encountered another issue related to cublasLtHSHMatmulAlgoInit. To address this, I uninstalled cublas from pip using the following command:
pip uninstall nvidia_cublas_cu11
After this, I successfully installed the DeepLM repository without any further issues [[100%] Built target BACore].

Anyway, I made several changes that may have influenced the outcome, as I initially encountered CUDA errors. For this reason, I will provide the complete pipeline I followed to successfully install DeepLM:

1. Modify the example.sh script, changing the line cmake .. to:
   cmake .. -DCMAKE_BUILD_TYPE=Release -DWITH_CUDA=ON -DCMAKE_CUDA_COMPILER=$(which nvcc)
2. Add to .bashrc
   export PATH="/usr/local/cuda-11.1/bin:$PATH" export LD_LIBRARY_PATH="/usr/local/cuda-11.1/lib64:$LD_LIBRARY_PATH" export TCNN_CUDA_ARCHITECTURE=86 export CUDA_HOME="/usr/local/cuda-11.1"
3. Check the GCC version (As I previously mentioned)
4. Run sh example.sh