OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-07 12:21:27 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity
94,812 Commits 8,304 Branches 150 Tags 1.5 GiB
5211f4c108
Commit Graph

8 Commits

Author SHA1 Message Date
Michael Suo
30fb2c4aba [lint] autoformat test/cpp and torch/csrc 
Let's have some fun.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/78828

Approved by: https://github.com/ezyang
 2022-06-11 21:11:16 +00:00
Bin Bao
25c6ebd12c Revert "Revert "[LT] Codegen ReuseNode for supported ops"" 
Summary: Fixed a XLC build failure by generating an always-return-false
default CanBeReused method.

This reverts commit 3cade9d454.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/77513

Approved by: https://github.com/alanwaketan
 2022-05-16 20:14:42 +00:00
PyTorch MergeBot
3cade9d454 Revert "[LT] Codegen ReuseNode for supported ops" 
This reverts commit 6066e5929f.

Reverted https://github.com/pytorch/pytorch/pull/76738 on behalf of https://github.com/malfet
 2022-05-14 00:33:10 +00:00
Bin Bao
6066e5929f [LT] Codegen ReuseNode for supported ops 
Summary:
1. Update the codegen script to add a TrieCache lookup (ReuseNode)
before creating a new IR node. The following is an example generated
code,

```
    at::Tensor LazyNativeFunctions::add(const at::Tensor & self, const at::Tensor & other, const at::Scalar & alpha) {
        ...
        torch::lazy::NodePtr node = torch::lazy::ReuseNode<AddTensor>(lazy_self->GetIrValue(), lazy_other->GetIrValue(), node_alpha);
        if (!node) {
            auto out_meta = at::meta::add(self, other, alpha);
            std::vector<Shape> shapes{Shape(out_meta.scalar_type(), out_meta.sizes().vec())};
            TORCH_INTERNAL_ASSERT(shapes.size() == 1);
            if(symbolicShapeEnabled()){
                std::vector<jit::IValue> inputs = { self, other, alpha };
                char* schema_str = "aten::add.Tensor(Tensor self, Tensor other, *, Scalar alpha=1) -> Tensor";
                applySymbolicShapesOnLT(schema_str, inputs, shapes);
            }

            node = torch::lazy::MakeNode<AddTensor>(lazy_self->GetIrValue(), lazy_other->GetIrValue(), node_alpha, std::move(shapes));
            CacheNode(node);
        }
        ...
    }
```
2. TrieCache lookup depends on each IR node subclass to provide its own
comparison function. The following is an example generated code,

```
  bool CanBeReused(const torch::lazy::Value& self, const torch::lazy::Value& other, const torch::lazy::Value& alpha) const {
    size_t i = 0;
    return (operand(i++) == self &&
        operand(i++) == other &&
        operand(i++) == alpha);
  }
```

3. DeviceData is specially handled.

4. Non-codegen op changes are coming a separate PR.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/76738

Approved by: https://github.com/JackCaoG, https://github.com/wconstab
 2022-05-13 19:13:58 +00:00
Bin Bao
8f5cdc6d5d Revert "Revert "[LT] Store OpKind for each IR subclass in a static field"" 
Summary: Re-land https://github.com/pytorch/pytorch/pull/76711 by
fixing internal build errors.
Generate class-level opkind as a static method instead of a static
member.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/77102

Approved by: https://github.com/wconstab, https://github.com/JackCaoG, https://github.com/antoniojkim
 2022-05-11 12:27:05 +00:00
PyTorch MergeBot
7eaf4780ba Revert "[LT] Store OpKind for each IR subclass in a static field" 
This reverts commit ac37ddc795.

Reverted https://github.com/pytorch/pytorch/pull/76711 on behalf of https://github.com/malfet
 2022-05-09 20:50:09 +00:00
Bin Bao
ac37ddc795 [LT] Store OpKind for each IR subclass in a static field 
Summary: Currently OpKind is stored as an object field called op_ for each IR
node, and one usage of op_ is to avoid dynamic_cast in NodeCast when we
need to downcast a base-node pointer into a concrete sub-node pointer.
As a result, we need to construct and pass in an op when downcasting
nodes, and this becomes quite anonnying when we start to implement the
trie-based IR node reusing. More importantly, the op for each subclass
should be unique for that subclass and thus making it a const static field
is a more logical design.

In this PR, we still keep the object-level op_ for easier XLA adoption. As
furture work, we can come back to remove op_, make the op() method
virtual, and get rid of OpKind in all the node constructors.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/76711

Approved by: https://github.com/wconstab, https://github.com/JackCaoG
 2022-05-06 19:14:46 +00:00
Bin Bao
f05710dd40 [LT] Add a trie data structure for caching IR nodes 
Summary: TrieCache provides a way to look up an IR node before we
actually create it. If the lookup hits in TrieCache, we reuse the
existing node and move the current pointer in TrieCache to point to that
node; if the lookup misses, we create a new node and insert it into TrieCache.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/76542

Approved by: https://github.com/wconstab, https://github.com/JackCaoG
 2022-05-04 23:48:03 +00:00