df.to_csv('./dataset.csv', index=False) crashing with "internal error" in RAPIDS 24.02 cuFile

[frasertajima]

Describe the bug
Running 02_Introduction_to_cuDF.ipynb notebook from Rapids sample notebooks results in cuFile "internal error":

File csv.pyx:545, in cudf._lib.csv.write_csv()

RuntimeError: cuFile error at: [/var/home/fraser/anaconda3/envs/rapids-24.02/include/kvikio/file_handle.hpp:178](https://file+.vscode-resource.vscode-cdn.net/var/home/fraser/anaconda3/envs/rapids-24.02/include/kvikio/file_handle.hpp:178): internal error

Fedora Silverblue 40 beta CUDA is 12.4 but the Conda environment installs CUDA 12.0:

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2022 NVIDIA Corporation
Built on Mon_Oct_24_19:12:58_PDT_2022
Cuda compilation tools, release 12.0, V12.0.76
Build cuda_12.0.r12.0/compiler.31968024_0

Steps/Code to reproduce bug

Run notebook cell 27:

df.to_csv('./dataset.csv', index=False)

the following notebook cell 28 also crashes with same error:

df = cudf.read_csv('./dataset.csv')

Expected behavior
Writing/reading of df dataframe to/from csv file.

Environment details (please complete the following information):

• Ubuntu 23.04 x86_64 distrobox container hosted by Fedora Silverblue 40 beta

• Kernel: 6.8.1-300.fc40.x86_64

• CUDA 12.0 from conda install

• Method of RAPIDS libraries install: [conda, Docker, or from source]

  • conda:
    conda create --solver=libmamba -n rapids-24.02 -c rapidsai -c conda-forge -c nvidia rapids=24.02 python=3.10 cuda-version=12.0

Additional context
This error has existed since January's rapids 23.12 when I last tried Rapids out. (cudf read csv)