Description: This seminar will cover how to build Miniforge environments and Singularity containers for TensorFlow and PyTorch on HPC systems with AMD GPUs. The AMD ROCM Dockerhub channel and AMD pip package repositories will be covered as well as other build differences for external tools like Horovod for TensorFlow. Examples will expand on previous documentation in the AI & ML User Guide to show how the same code can be deployed on systems with MI250 and MI300a GPUs. Differences between systems with Nvidia GPUs and AMD GPUs will be highlighted.