ROCm
An open-source GPU computing platform for AI training and inference on AMD hardware, with Linux and Windows support and PyTorch, JAX, vLLM and SGLang.
ROCm is AMD's open-source GPU computing platform for developers running AI training, inference and scientific workloads on their own hardware or servers. It supports selected Linux and Windows configurations on AMD Instinct, Radeon and Ryzen AI devices. Check the version-specific GPU, operating-system, driver and firmware compatibility matrix before installing. It's the software foundation for applications that need AMD GPU acceleration, including local LLM workloads.
Its AI ecosystem supports PyTorch and JAX, with vLLM and SGLang for inference. Developers can use these frameworks alongside GPU math libraries and communication libraries for workloads that span multiple GPUs. ROCm also supports TensorFlow.
For developers writing GPU code, HIP provides a C++ programming interface similar to NVIDIA CUDA and supports portable GPU code. ROCm also supports OpenMP and OpenCL. The Core SDK brings together compilers, runtimes and compute libraries, plus profiling and debugging tools for finding performance bottlenecks and diagnosing failures.
ROCm covers deployments beyond a single workstation. Its infrastructure tools support GPU monitoring, partitioning, virtualization and containers, with Kubernetes integration through the Network Operator. The platform also supports cloud deployments, so where a workload runs depends on the infrastructure you choose.
Domain-specific toolkits include ROCm Data Science, Finance, Life Science, LLMExt and Simulation. Components have their own licenses. The MIT-licensed legacy-rocm-build repository directs new build work to ROCm/TheRock.
