Installation Error for Gsplat

[FudongGe]

When trying to "pip install git+https://github.com/nerfstudio-project/gsplat.git@v1.3.0", I meet this problem. Did someone know how to fix it?

[ziyc]

Hi @FudongGe,

This might be a CUDA-PyTorch version mismatch. Have you installed the correct PyTorch version for your CUDA? We support gsplat>=1.0.0. You can check the gsplat repo for similar issues and solutions.

I hope this response is helpful. If you have more questions, feel free to ask.

[FudongGe]

Hi @FudongGe,

This might be a CUDA-PyTorch version mismatch. Have you installed the correct PyTorch version for your CUDA? We support gsplat>=1.0.0. You can check the gsplat repo for similar issues and solutions.

I hope this response is helpful. If you have more questions, feel free to ask.

Hi, Ziyu.

Thanks for your response.

Actually, I follow your installation guide and install torch==2.0.0+cu117. My CUDA toolkit version is 11.7, and my CUDA driver can support up to 12.2. I don't think this will lead to version conflicts. Do you have any other ideas？

[ziyc]

emmmm, it's strange, can you try pip install gsplat==1.3.0 and just run? If it doesn't work, you could try other versions like pip install gsplat==1.0.0, if it still got error, feel free to show it to me.

[FudongGe]

emmmm, it's strange, can you try pip install gsplat==1.3.0 and just run? If it doesn't work, you could try other versions like pip install gsplat==1.0.0, if it still got error, feel free to show it to me.

Yes, direct 'pip install gsaplt==1.3.0' is ok. But about whether it can run successfully, I will try. Thanks. By the way, what's the difference between the two installation methods ? Thanks again!

[ziyc]

Hi @FudongGe, these two ways determine when the lib is compiled. As described here:

[FudongGe]

Hi @FudongGe, these two ways determine when the lib is compiled. As described here:

Thanks very much！