Implement aten::{all,any}.dims | feat(torchlib) #1084
Conversation
justinchuby
added
the
topic: torch_lib
Codecov Report
@@ Coverage Diff @@
## main #1084 +/- ##
==========================================
- Coverage 78.10% 77.98% -0.13%
==========================================
Files 115 115
Lines 14826 14860 +34
Branches 1574 1580 +6
==========================================
+ Hits 11580 11588 +8
- Misses 2875 2901 +26
Partials 371 371
|
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
Test Results 18 files ± 0 18 suites ±0 1h 18m 35s ⏱️ + 5m 14s For more details on these failures, see this check. Results for commit 0ccd9ad. ± Comparison against base commit a981b8a. ♻️ This comment has been updated with latest results. |
|
|
||
|
|
||
| @torch_op("aten::any.dims") | ||
| def aten_any_dims_empty_dim(self: TTensor, keepdims: bool) -> BOOL: |
There was a problem hiding this comment.
Looks like we still need the place holder for it when it is optional
There was a problem hiding this comment.
NVM. dims is an attribute. This is good.
|
|
||
|
|
||
| @torch_op("aten::any.dims") | ||
| def aten_any_dims_empty_dim(self: TTensor, keepdims: bool) -> BOOL: |
There was a problem hiding this comment.
nit:
Change 'empty' to 'no' to avoid confusion that users might think dims is an empty list/tuple where it is actually None.
There was a problem hiding this comment.
dims can be an empty list or None. They will mean the same thing. I made the changes nevertheless.
Implement aten::{all,any}.dims according to pytorch/pytorch#110310
Tests will be enabled after the PyTorch PR is merged