cudaDriverEntryPointQueryResult not defined

[aviczhl2]

During installation (from main branch with pip install .), it reports error: ‘cudaDriverEntryPointQueryResult’ was not declared in this scope

on line 98 of transformer_engine/common/util/cuda_driver.cpp
98 | cudaDriverEntryPointQueryResult driver_result;

I try to search for dependency and It seems that this is not defined anywhere in this project. Where is it defined and how to solve the problem? Thanks.

I'm using python 3.12.4, cuda-11.8 and pytorch=2.4.0

[timmoon10]

Can you try upgrading to CUDA 12.0 or newer? See #1086 (comment).

[ptrendx]

@timmoon10 Could you update our installation instructions? We are still listing 11.8 as the prerequisite here: https://docs.nvidia.com/deeplearning/transformer-engine/user-guide/installation.html

[aviczhl2]

Yeah the problem was solved when I use CUDA-12.2