Warning to `Install tcnn for speedups` even when tcnn is installed

[Ben-Epstein]

Describe the bug This may not be a bug, rather asking for help debugging or more clear warning messages.

I'm running the training process via ns-train nerfacto command, and I keep seeing the warnings

WARNING: Using a slow implementation for the SHEncoding module. 
🏃 🏃 Install tcnn for speedups 🏃 🏃
pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch


WARNING: Using a slow implementation for the NeRFEncoding module. 
🏃 🏃 Install tcnn for speedups 🏃 🏃
pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch


WARNING: Using a slow implementation for the MLPWithHashEncoding module. 
🏃 🏃 Install tcnn for speedups 🏃 🏃
pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch


WARNING: Using a slow implementation for the MLP module. 
🏃 🏃 Install tcnn for speedups 🏃 🏃
pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch


WARNING: Using a slow implementation for the HashEncoding module. 
🏃 🏃 Install tcnn for speedups 🏃 🏃
pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch


WARNING: Using a slow implementation for the MLP module. 
🏃 🏃 Install tcnn for speedups 🏃 🏃
pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch


WARNING: Using a slow implementation for the HashEncoding module. 
🏃 🏃 Install tcnn for speedups 🏃 🏃
pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch


WARNING: Using a slow implementation for the MLP module. 
🏃 🏃 Install tcnn for speedups 🏃 🏃
pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch

Even though I have installed it, and can verify it in my pip modules.

I installed it with pip install 'git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch' and didn't see any errors on installation.

pip version 23.3.1, python vesion 3.10.9, Ubuntu 22.04.3 LTS

Any thoughts what might be happening? Thanks!

[jb-ye]

Can you try import tinycudann as tcnn from your python console? It might be possible that you have multiple python environments and the one lives with nerfstudio doesn't have tcnn.

[Daozuqi]

yeh, 'import tinycudann as tcnn' works but I still get that warning. besides, everytime I activate the env, it will show "tinycudann is not installed! Please install it for faster training.". I dont have another env to see whether tinycudann is working or not. If you have solutions, please share with me. thanks

[kevinycy]

You can build it locally to solve this problem, by https://github.com/NVlabs/tiny-cuda-nn. With $ git clone --recursive https://github.com/nvlabs/tiny-cuda-nn $ cd tiny-cuda-nn tiny-cuda-nn$ cmake . -B build -DCMAKE_BUILD_TYPE=RelWithDebInfo tiny-cuda-nn$ cmake --build build --config RelWithDebInfo -j tiny-cuda-nn$ cd bindings/torch tiny-cuda-nn/bindings/torch$ python setup.py install