3 Commits

Author	SHA1	Message	Date
Teja	dd3e7170c2	Add async checkpointing impl to experimental checkpointer and add a builder API (#156927) ... 1. Adds an AsyncCheckpointer with out-of-process checkpointing and state_dict_stager with shared memory, pinned memory and Zero Overhead Support. 2. Adds two conveinient functions to create sync/async checkpointers Differential Revision: [D77336833](https://our.internmc.facebook.com/intern/diff/D77336833/) Pull Request resolved: https://github.com/pytorch/pytorch/pull/156927 Approved by: https://github.com/pradeepfn	2025-07-03 22:49:20 +00:00
Teja	ef6dfa06a9	Create a base Checkpointer and SyncCheckpointer and add dist barrier impl and (#156926) ... In preparation to adding async checkpointing, this diff adds 1. Change Checkpointer to an Abstract base class and adds a sync checkpointer implementation. 2. torch.distributed.barrier() as one of the barrier choices. Differential Revision: [D77341314](https://our.internmc.facebook.com/intern/diff/D77341314/) Pull Request resolved: https://github.com/pytorch/pytorch/pull/156926 Approved by: https://github.com/pradeepfn	2025-06-28 02:48:29 +00:00
Teja Rao	d797038ea9	[dcp_poc] Introduce a new simple rank local checkpointer (#156142) ... Summary: Adds an experimental implementation for a rank local checkpointer with save and load with partial load, blind load and in-place load. This uses an new API and simpler format. Plan to add async checkpointing, IO layer, pluggable storage backend, layout customization, Resharding, deduplication etc are not implemented. Test Plan: unit tests Reviewed By: saumishr Differential Revision: D75426560 Pull Request resolved: https://github.com/pytorch/pytorch/pull/156142 Approved by: https://github.com/saumishr	2025-06-25 01:19:40 +00:00