ROCm
An open-source GPU computing platform for local AI training and inference on AMD hardware, with Linux and Windows support and PyTorch, JAX and vLLM.
ROCm is AMD's GPU computing platform for developers running AI training, inference and scientific workloads on their own workstations or servers. It provides the software needed to use AMD GPUs for these workloads, including local LLM inference.
- AI frameworks and inference: Run workloads with PyTorch, JAX and TensorFlow, or use vLLM and SGLang for model inference.
- GPU programming: HIP lets developers write portable GPU code in C++ with an interface similar to NVIDIA CUDA. The platform also supports OpenMP and OpenCL.
- Compute and performance tools: Math and communication libraries support GPU workloads, while profiling and debugging tools help identify bottlenecks and diagnose failures.
- Server and cluster deployments: Infrastructure tools support GPU monitoring, partitioning, virtualization and containers, with Kubernetes integration through the Network Operator.
ROCm is open source and supports Linux and Windows on compatible AMD Instinct, Radeon and Ryzen AI devices. Hardware and operating-system support depends on the ROCm compatibility matrix, so support for one AMD device doesn't imply support for every GPU or platform combination.
Workloads can run on your own hardware or in cloud infrastructure. The MIT-licensed legacy-rocm-build repository points build development to ROCm/TheRock.
Logotyper
