Revision history [back]

If you are encountering an error while attempting to utilize onnxruntime with GPU by using SessionOptionsAppendExecutionProvider_CUDA, there could be several causes and possible solutions. Here are a few things you can try:

1. Make sure that CUDA drivers and libraries are properly installed on your system. You can do this by running the CUDA sample programs included with the CUDA toolkit.

2. Verify that the CUDA version you have installed on your system is compatible with the version of the NVIDIA GPU you are using. Sometimes, mismatched versions can cause issues.

3. Ensure that your system has enough GPU memory available to run your model. If the model is too large to fit in GPU memory, you may need to split it up into smaller parts or use a different execution provider.

4. Check that you're using the latest version of onnxruntime and that it's correctly installed. If not, you may need to update or reinstall it.

5. If you're running on Windows, try running your code in administrator mode. This can help resolve permission issues that may be preventing the GPU from being accessed.

6. Check the logs or error messages to identify the specific issue and investigate further. There may be additional steps you need to take depending on the error message.

If none of these solutions work, you may need to seek additional help in forums or support channels specific to onnxruntime or your GPU manufacturer.