research-pytorch-fsdp2
Adds PyTorch FSDP2 (fully_shard) to training scripts with correct init, sharding, mixed precision/offload config, and distributed checkpointing. Use when models exceed single-GPU memory or when you...
来源信息
- 仓库
- GrayCodeAI/starling
- 最近来源活动
- 2026年7月13日 09:28
- 检测到的 SKILL.md 语言
- 英语
- 星标
- 2
- 分支
- 0
安装方式
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
检查来源文件
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
正在显示 SKILL.md
SKILL.md
来源说明 · 只读预览- name
- research-pytorch-fsdp2
- description
- Adds PyTorch FSDP2 (fully_shard) to training scripts with correct init, sharding, mixed precision/offload config, and distributed checkpointing. Use when models exceed single-GPU memory or when you...
- license
- MIT
- tags
- ["pytorch","fsdp2","fully-sharded-data-parallel","distributed-training","dtensor"]
- version
- 1.0.0
- author
- Orchestra Research
- dependencies
- ["torch"]
## References
- `references/pytorch_fsdp2_tutorial.md`
- `references/pytorch_fully_shard_api.md`
- `references/pytorch_ddp_notes.md`
- `references/pytorch_fsdp1_api.md`
- `references/pytorch_device_mesh_tutorial.md`
- `references/pytorch_tp_tutorial.md`
- `references/pytorch_dcp_overview.md`
- `references/pytorch_dcp_recipe.md`
- `references/pytorch_dcp_async_recipe.md`
- `references/pytorch_examples_fsdp2.md`
- `references/torchtitan_fsdp_notes.md` (optional, production notes)
- `references/ray_train_fsdp2_example.md` (optional, integration example)
在 GitHub 查看