[ Solution ] For error install "fused-bilagrid" & "fused-ssim" ["/example/requirements.txt"]

[ToonKru3]

GPU Environment Setup Troubleshooting

Common GCC/G++ Version Compatibility Issue

When installing Python 3.10, PyTorch, and CUDA, you may encounter GCC & G++ version compatibility errors during package installation.

Error Symptoms

If you see an error message like this:

#error -- unsupported GNU version! GCC versions later than 11 are not supported! 
The nvcc flag '-allow-unsupported-compiler' can be used to override this version check. 
However, using an unsupported host compiler may cause compilation failure or incorrect run-time execution.

This indicates that packages like "fused-bilagrid" and "fused-ssim" require GCC/G++ versions < 11. Even versions like 11.x will cause compilation errors.

Solution 1: Override Compiler Checks

Set the following environment variables to allow unsupported compiler versions:

Depending on your Hardware
export TORCH_CUDA_ARCH_LIS="version"
python -c "import torch; print(torch.cuda.get_device_capability())"

export TORCH_CUDA_ARCH_LIST="8.6"
export CXX=g++
export CC=gcc
export CUDAFLAGS="--allow-unsupported-compiler"

Then retry the installation:

pip install -r requirements.txt

Solution 2: Install Compatible GCC/G++ Version (Recommended)

Install GCC and G++ version 10, which is known to be compatible:

sudo apt install -y gcc-10 g++-10
export CC=/usr/bin/gcc-10
export CXX=/usr/bin/g++-10

Then retry the installation:

pip install -r requirements.txt

Notes

• Solution 2 is generally more reliable as it uses officially supported compiler versions
• Make sure to set the environment variables in the same terminal session where you run the pip install command

[BigyanBhatta]

Clone "fused-bilagrid" and "fused-ssim" manually and while installing requirements, run this command which prevents installing already installed packages:

pip install -r examples/requirements.txt --no-deps