Tag
在 PyTorch Conference North America 2026 上，AMD 团队将展示 MoRI 通信栈与 vLLM 结合，实现跨 pod 的宽专家并行（Wide Expert Parallelism）以及基于 RDMA 的 KV 缓存传输，用于 AMD Instinct GPU 上的分离式 MoE 服务。
The article promotes a poster presentation at PyTorch Conference North America on enabling the open-source vime RL post-training framework on AMD Instinct GPUs using ROCm, and provides registration details for the conference.
Built an open-source pipeline that takes a single sentence and produces a cinematic reel with characters, animation, music, and narration, using FLUX.2, Wan2.2, and other models on a single AMD GPU. The pipeline includes a director agent, character generation, keyframe animation, vision critic, music, and narration stages.