PyTorch Monarch Now Supports AMD GPUs for Distributed Training

PyTorch has announced that its Monarch framework is now compatible with AMD GPUs. The integration leverages the ROCm software stack for hardware acceleration. Users can perform single‑controller distributed training across AMD devices. The update expands PyTorch’s hardware ecosystem beyond NVIDIA GPUs. Documentation provides guidance for configuring Monarch on AMD platforms. Early benchmarks show comparable scaling efficiency to NVIDIA setups. The community is expected to adopt the feature for diverse AI workloads. Future releases may further optimize performance and add more AMD‑specific tools.