英伟达开源 Reshard 集合通信，强化学习权重 refit 提速最高 5.7 倍
![[object Object]](https://developer-blogs.nvidia.com/wp-content/uploads/2026/11/green-cube.webp)
NCCL M2N Reshard 只传输相交分片；NeMo RL 基准相对 gather-and-broadcast 提速 2.7–5.7 倍。
为何重要
英伟达介绍 Reshard：将强化学习权重 refit 视为 DTensor 到 DTensor 传输，只搬迁理论最小相交分片。
要点
- New Reshard collective for DTensor-to-DTensor refit
- Open NCCL M2N Reshard reference implementation
- 2.7×–5.7× faster than gather-and-broadcast in NeMo RL
- Benchmarked on DGX H100 and GB200
- Hybrid path: FFN projections via Reshard; small weights legacy
- Aimed at trillion-parameter RL post-training payloads