Commit c8b02b6
authored
Ensured that output tensors (output and indices) correctly propagate the names from the input tensor (#2281)
Fix #2023
Makes improvements to the `max_pool3d_with_indices_xpu` implementation
by enhancing support for named tensors and ensuring proper propagation
of tensor names. The changes also include a minor code organization
update.
Enhancements for named tensor support:
* Added an include for `ATen/NamedTensorUtils.h` in
`DilatedMaxPool3d.cpp` to enable named tensor utilities.
* Ensured that output tensors (`output` and `indices`) correctly
propagate the names from the input tensor by calling
`namedinference::propagate_names` for both.1 parent fd598b0 commit c8b02b6
1 file changed
+5
-0
lines changed| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
1 | 2 | | |
2 | 3 | | |
3 | 4 | | |
| |||
17 | 18 | | |
18 | 19 | | |
19 | 20 | | |
| 21 | + | |
20 | 22 | | |
21 | 23 | | |
22 | 24 | | |
| |||
26 | 28 | | |
27 | 29 | | |
28 | 30 | | |
| 31 | + | |
| 32 | + | |
| 33 | + | |
29 | 34 | | |
30 | 35 | | |
31 | 36 | | |
| |||
0 commit comments