mirror of
https://github.com/pytorch/pytorch.git
synced 2026-11-20 09:04:25 +08:00
19da3da3d3
The fix-cpp-build branch is a major restructuring that: - Drops Exhale (which auto-generated API docs but caused OOM in CI) in favor of hand-written RST pages using Breathe directives - Reorganizes notes into a hierarchical API reference with proper categories - Adds new content: XPU docs, stable ABI docs, DeviceGuard/Stream docs - Always runs Doxygen instead of gating on an env var Pull Request resolved: https://github.com/pytorch/pytorch/pull/174096 Approved by: https://github.com/janeyx99, https://github.com/seemethere, https://github.com/albanD
1.4 KiB
1.4 KiB
myst
| myst | ||||||
|---|---|---|---|---|---|---|
|
Serialization (torch::serialize)
The torch::serialize namespace provides utilities for saving and loading
model weights, tensors, and optimizer state. This enables checkpointing during
training and deployment of trained models.
When to use torch::serialize:
- When saving trained models to disk
- When implementing training checkpoints
- When loading pre-trained weights
- When transferring models between C++ and Python
Basic usage:
#include <torch/torch.h>
// Save a model
auto model = std::make_shared<Net>();
// ... train the model ...
torch::save(model, "model.pt");
// Load a model
auto loaded_model = std::make_shared<Net>();
torch::load(loaded_model, "model.pt");
// Save and load tensors
torch::Tensor tensor = torch::randn({2, 3});
torch::save(tensor, "tensor.pt");
torch::Tensor loaded_tensor;
torch::load(loaded_tensor, "tensor.pt");
Header Files
torch/csrc/api/include/torch/serialize.h- Main serialization headertorch/csrc/api/include/torch/serialize/archive.h- Archive classestorch/csrc/api/include/torch/serialize/input-archive.h- Input archivetorch/csrc/api/include/torch/serialize/output-archive.h- Output archive
Serialization Categories
:maxdepth: 1
save_load
archives
checkpoints