Fix device placement of constants, shared inputs and delegate outputs by shoumikhin · Pull Request #24088 · pytorch/executorch

Normalize the dim order of parameters before delegation, like constants

shoumikhin

committed

0a7d3f9

Keep a device round-trip when a host op writes the host copy

shoumikhin

committed

4c41bca

Copy a device delegate output to the host when it is used without getitem

shoumikhin

committed

d95d1f6

Return delegate outputs on the device whenever skip_d2h_for_method_outputs is set

shoumikhin

committed

6e48878

Copy constants that feed a device delegate into device memory

shoumikhin

committed

019f0eb

Let a method input feed several delegates on one device without a copy

shoumikhin

committed

c11f648

Copy a tensor to a device once when several delegates there read it

shoumikhin

committed

d9c4144

Keep in-place results on the host copy of a device delegate output

shoumikhin

committed

9f5990a

Count a host write through a view as a write to the tensor

shoumikhin

committed

19f9279

Test tensors shared by delegates on different devices

shoumikhin

committed

cab98dd

Test device placement end to end with CUDA delegates

shoumikhin

committed

23aa8e9

Say when skip_d2h_for_method_outputs still copies a returned output

shoumikhin

committed

f08b165

Ignore host writes that come after a device copy is read

shoumikhin

committed

a0bbb9d

Share the device copy of a written output with a later delegate

shoumikhin

committed

ed70bc5

Check the placeholder, not the input spec, in convert_constant_dim_order_pass

shoumikhin

committed

3693aea

Run the CUDA unit tests when the device placement passes change

shoumikhin

committed

0ac8098

Navigation Menu

Repository navigation

Fix device placement of constants, shared inputs and delegate outputs - #24088

Fix device placement of constants, shared inputs and delegate outputs#24088
shoumikhin wants to merge 16 commits into
pytorch:mainfrom
shoumikhin:device-pass-constants

Commits on Oct 24, 2026

Normalize the dim order of parameters before delegation, like constants

Keep a device round-trip when a host op writes the host copy

Copy a device delegate output to the host when it is used without getitem

Return delegate outputs on the device whenever skip_d2h_for_method_outputs is set

Copy constants that feed a device delegate into device memory

Let a method input feed several delegates on one device without a copy

Copy a tensor to a device once when several delegates there read it

Keep in-place results on the host copy of a device delegate output

Count a host write through a view as a write to the tensor

Test tensors shared by delegates on different devices

Test device placement end to end with CUDA delegates

Say when skip_d2h_for_method_outputs still copies a returned output

Ignore host writes that come after a device copy is read

Share the device copy of a written output with a later delegate

Check the placeholder, not the input spec, in convert_constant_dim_order_pass

Run the CUDA unit tests when the device placement passes change