The AMD Developer Cloud provides developers with easy access to powerful AMD Instinct GPUs through a streamlined platform offering free initial cloud credits, flexible VM configurations, and pre-configured development environments with popular AI frameworks. Users can quickly set up GPU VMs, access tools like JupyterLab with tutorials, and start accelerating their AI and computational projects without complex setup or hardware investment.
The AMD Developer Cloud is a developer-centric platform designed to provide seamless and easy access to AMD Instinct GPUs, enabling fast and flexible development without the need for complex setup. The platform offers powerful performance and is tailored to help developers get started quickly with GPU computing. The video guides users through the process of setting up an account, requesting credits, and launching GPU virtual machines (VMs) to begin their development work within minutes.
To start using the AMD Developer Cloud, users need to sign up by creating an account using either their email or GitHub ID. AMD offers an initial 25 hours of complimentary cloud credits to qualified developers who apply and are accepted, allowing them to access AMD GPUs without upfront hardware investment. Developers can request these credits by filling out a form, and if approved, the credits are added to their account. Those not approved can still access GPUs via a pay-as-you-go model, subject to terms and conditions.
Once the account and credits are set up, users can create a GPU VM by selecting either a single MI300X GPU or a full node with eight GPUs. They then choose a base image for their VM, which can be a bare operating system, custom images, or pre-configured images that include ROCm and other development tools. The platform offers Quick Start packages with Docker images configured for popular frameworks such as vLLM, S G Lang, PyTorch, Megatron, and JAX, each accompanied by detailed guides and version information.
The video highlights the vLLM Quick Start image as an example, which includes a base ROCm system, an official vLLM Docker container optimized for ROCm, and JupyterLab accessible directly from the browser. This setup comes with five preloaded tutorials that users can run within JupyterLab, providing hands-on experience with AI development tools. Users upload their SSH keys during VM creation and can access their GPU VM through SSH, a web console, or JupyterLab, making it easy to verify GPU availability and start working immediately.
Finally, the video demonstrates accessing JupyterLab by copying the VM’s IP address, retrieving an access token, and logging in to run tutorials. One example tutorial shows how to build an AI agent using MCP tools, vLLM, Pydantic AI, and the AMD MI300X GPU. With these resources and tools readily available, developers are fully equipped to begin their projects on the AMD Developer Cloud, leveraging powerful GPU resources and streamlined workflows to accelerate their AI and computational workloads.