kermt-setup

To set up a Docker environment for BioNeMo workflows, you need to verify host system requirements, build the specialized containerized image, and validate CUDA-capable GPU passthrough for downstream model execution. This skill automates that entire configuration process.

Running CUDA-based scientific computing containers requires Docker, the NVIDIA Container Toolkit, and CUDA-capable GPU hardware. The environment validation process automatically checks your host system for these specific dependencies before attempting to build the execution image.

To verify PyTorch and CUDA connectivity inside a container, you can run an internal GPU smoke test. This validates that the GPU passthrough is functional and that the PyTorch framework successfully communicates with the CUDA hardware before starting scientific workloads.

Yes, the NVIDIA Container Toolkit is strictly required to build BioNeMo agent environments. The setup process depends on it alongside Docker and functional GPU drivers to enable the necessary hardware passthrough for downstream model execution.

BioNeMo environment setup fails during the Docker image build if host system requirements are unmet. The automated validation checks for Docker, the NVIDIA Container Toolkit, and CUDA-capable GPU availability to isolate missing dependencies preventing the image construction.