frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-20 21:14:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

[ca] side-effect free inital trace: compiled_args (#147804)

Browse Source 
const methods to prevent accidental mutation. changes mainly in Error nodes and PyNode.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/147804
Approved by: https://github.com/jansel
ghstack dependencies: #147242, #147796
This commit is contained in:
Simon Fan
2025-02-25 19:57:55 -08:00
committed by PyTorch MergeBot
parent 5e3069dde8
commit fd1220e386
17 changed files with 105 additions and 79 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
torch/csrc/autograd/function.h
View File

@ -545,8 +545,8 @@ struct TORCH_API Node : std::enable_shared_from_this<Node> {
return tensor_pre_hooks_;
}
virtual std::unique_ptr<PostAccumulateGradHook>&
tensor_post_acc_grad_hooks() noexcept {
virtual std::unique_ptr<PostAccumulateGradHook>& tensor_post_acc_grad_hooks()
const noexcept {
static std::unique_ptr<PostAccumulateGradHook> empty = nullptr;
return empty;
}
@ -593,7 +593,7 @@ struct TORCH_API Node : std::enable_shared_from_this<Node> {
// 2) Collect node information for specialization and caching
// Implementations in subclasses should call args.collect() with all node
// attrs. These functions are only called durring backward.
virtual void compiled_args(CompiledNodeArgs& args) {
virtual void compiled_args(CompiledNodeArgs& args) const {
throw std::runtime_error(
std::string("compiled_args not implemented: ") + name());
}