@BWise As a first step in troubleshooting this issue, please configure your worker and test the failed job with CUDA-11.2: