DLI Acclerated data science workflows self paced course giving CUDF error

AI & Data Science Deep Learning (Training & Inference) cuDNN
1

When running the following command in Jupyter :
%time gdf = cudf.read_csv(‘./data/pop_1-03.csv’)
gdf.shape

gives the following error:
missing cuda symbols while dynamic loading cuFile initialization failed CPU times: user 2.41 s, sys: 585 ms, total: 3 s Wall time: 2.98 s
(58479894, 6)

Please advise how to resolve this error.

4

Hi @seemant_mathur ,
Can you pls check on below pointers

  1. Check cuDNN Installation: Verify that cuDNN is properly installed on your system and that the correct version is compatible with your GPU drivers.
  2. Verify Environment Variables: Ensure that the environment variables related to cuDNN are correctly set. Check the paths for CUDA and cuDNN libraries in the system environment variables.
  3. Restart Jupyter Kernel: Sometimes, simply restarting the Jupyter kernel can resolve issues related to library dependencies. Try restarting your Jupyter notebook.
  4. Update cuDNN and CUDA Drivers: Check for updates to both the cuDNN library and CUDA drivers. Updating to the latest versions can fix compatibility issues.
  5. Check System Logs: Look for any error messages or warnings in the system logs that may provide more details about the error you are encountering. Both Jupyter logs and system logs should be checked.
  6. Reinstall cuDNN: If the problem persists, consider reinstalling cuDNN. Remove the existing installation and perform a fresh installation following the official documentation provided by NVIDIA.