reyoung / torchsnapshot

A performant, memory-efficient checkpointing library for PyTorch applications, designed with large, complex distributed workloads in mind.

torchsnapshot

This library is currently in Alpha and currently does not have a stable release. The API may change and may not be backward compatible. If you have suggestions for improvements, please open a GitHub issue. We'd love to hear your feedback.

A performant, memory-efficient checkpointing library for PyTorch applications, designed with large, complex distributed workloads in mind.

Install

Requires Python >= 3.7 and PyTorch >= 1.12

From pip:

pip install --pre torchsnapshot-nightly

From source:

git clone https://github.com/pytorch/torchsnapshot
cd torchsnapshot
pip install -r requirements.txt
python setup.py install

Why TorchSnapshot

Performance

• TorchSnapshot provides a fast checkpointing implementation employing various optimizations, including zero-copy serialization for most tensor types, overlapped device-to-host copy and storage I/O, parallelized storage I/O.
• TorchSnapshot greatly speeds up checkpointing for DistributedDataParallel workloads by distributing the write load across all ranks (benchmark).
• When host memory is abundant, TorchSnapshot allows training to resume before all storage I/O completes, reducing the time blocked by checkpoint saving.

Memory Usage

• TorchSnapshot's memory usage adapts to the host's available resources, greatly reducing the chance of out-of-memory issues when saving and loading checkpoints.
• TorchSnapshot supports efficient random access to individual objects within a snapshot, even when the snapshot is stored in a cloud object storage.

Usability

• Simple APIs that are consistent between distributed and non-distributed workloads.
• Out of the box integration with commonly used cloud object storage systems.
• Automatic resharding (elasticity) on world size change for supported workloads (more details).

Security

• Secure tensor serialization without pickle dependency [WIP].

Getting Started

from torchsnapshot import Snapshot

# Taking a snapshot
app_state = {"model": model, "optimizer": optimizer}
snapshot = Snapshot.take(app_state=app_state, "/path/to/snapshot")

# Restoring from a snapshot
snapshot.restore(app_state=app_state)

See the documentation for more details.

License

torchsnapshot is BSD licensed, as found in the LICENSE file.