Expert guidance for Fully Sharded Data Parallel training with PyTorch FSDP - parameter sharding, mixed precision, CPU offloading, FSDP2. Use when training large models with native PyTorch distributed training support.
/plugin marketplace add zechenzhangAGI/AI-research-SKILLs/plugin install pytorch-fsdp@zechenzhangAGI/AI-research-SKILLs