Revision history [back]

To resolve this issue, you can follow these steps:

1. Check if the libnvrtc.so file is present in your system by running the following command:

   sudo find / -name 'libnvrtc.so'
   

   This will search for the file in your entire system. If the file is found, note down its path.

2. If the file is not found, you may need to install the NVIDIA CUDA Toolkit, which includes the libnvrtc.so file. Follow the installation instructions for your Linux distribution and make sure to include the CUDA toolkit.

3. Once you have located the libnvrtc.so file, make sure it is included in your LDLIBRARYPATH environment variable by running the following command:

   export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/path/to/libnvrtc.so
   

   Replace "/path/to/libnvrtc.so" with the path where the file is located.

4. Finally, try running your Torch application again. It should be able to locate the libnvrtc.so file and load the libcudnncnninfer.so.8 library without any issues.