OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-06 12:20:52 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity

[Intel GPU] Support SparseCsrXPU codegen (#144722)

Browse Source 
Adding a new dispatch key - `SparseCsrXPU`  to enable Intel GPU support for SparseCsr Tensor.

Similar PR: https://github.com/pytorch/pytorch/pull/139267
Pull Request resolved: https://github.com/pytorch/pytorch/pull/144722
Approved by: https://github.com/EikanWang, https://github.com/guangyey, https://github.com/albanD

Co-authored-by: Kanya-Mo <kanya.mo@intel.com>
This commit is contained in:
cfgfung 2025-02-16 03:16:12 +00:00 committed by PyTorch MergeBot
parent 1677a31019
commit 8f20026bcb
3 changed files with 6 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
aten/src/ATen/SparseCsrTensorImpl.cpp
View File

@ -56,9 +56,11 @@ SparseCsrTensorImpl::SparseCsrTensorImpl(
TORCH_INTERNAL_ASSERT(((key_set.has(DispatchKey::SparseCsrCPU) && device().type() == kCPU) TORCH_INTERNAL_ASSERT(((key_set.has(DispatchKey::SparseCsrCPU) && device().type() == kCPU)
|| (key_set.has(DispatchKey::SparseCsrCUDA) && device().type() == kCUDA) || (key_set.has(DispatchKey::SparseCsrCUDA) && device().type() == kCUDA)
|| (key_set.has(DispatchKey::SparseCsrXPU) && device().type() == kXPU)
|| (key_set.has(DispatchKey::SparseCsrMeta) && device().type() == kMeta) || (key_set.has(DispatchKey::SparseCsrMeta) && device().type() == kMeta)
|| (key_set.has(DispatchKey::SparseCsrCPU) && device().type() == kMeta) // fake tensor || (key_set.has(DispatchKey::SparseCsrCPU) && device().type() == kMeta) // fake tensor
|| (key_set.has(DispatchKey::SparseCsrCUDA) && device().type() == kMeta) // fake tensor || (key_set.has(DispatchKey::SparseCsrCUDA) && device().type() == kMeta) // fake tensor
|| (key_set.has(DispatchKey::SparseCsrXPU) && device().type() == kMeta) // fake tensor
|| (key_set.has(DispatchKey::SparseCsrPrivateUse1) && device().type() == kPrivateUse1)), || (key_set.has(DispatchKey::SparseCsrPrivateUse1) && device().type() == kPrivateUse1)),
"Inconsistent key_set (=", key_set, ") and device (=", device(), ")"); "Inconsistent key_set (=", key_set, ") and device (=", device(), ")");

3
aten/src/ATen/native/sparse/SparseCsrTensor.cpp
View File

@ -360,6 +360,9 @@ static SparseCsrTensor new_compressed_tensor(const TensorOptions& options) {
case kCUDA: case kCUDA:
dispatch_key = DispatchKey::SparseCsrCUDA; dispatch_key = DispatchKey::SparseCsrCUDA;
break; break;
case kXPU:
dispatch_key = DispatchKey::SparseCsrXPU;
break;
case kMeta: case kMeta:
dispatch_key = DispatchKey::SparseCsrMeta; dispatch_key = DispatchKey::SparseCsrMeta;
break; break;

1
torchgen/model.py
View File

@ -283,6 +283,7 @@ dispatch_keys = [
DispatchKey.MPS, DispatchKey.MPS,
DispatchKey.XPU, DispatchKey.XPU,
DispatchKey.SparseXPU, DispatchKey.SparseXPU,
DispatchKey.SparseCsrXPU,
DispatchKey.SparseCUDA, DispatchKey.SparseCUDA,
DispatchKey.SparseCsrCUDA, DispatchKey.SparseCsrCUDA,
DispatchKey.QuantizedCPU, DispatchKey.QuantizedCPU,