ehsanmok commented on code in PR #13163:
URL: https://github.com/apache/tvm/pull/13163#discussion_r1002843613
##########
tests/python/frontend/onnx/test_forward.py:
##########
@@ -5383,6 +5375,12 @@ def verify_eyelike(indata, dynamic=False):
]
}
+flaky_target_skips = {
Review Comment:
Unfortunately not, bc it fails if cuda is enabled but passes without it
locally. Here's the minor differences between the actual onnx val and tvm val
for the `gridsample_nearest` test case on nvptx
```
actual = array([[[[0., 0., 2., 2.],
[2., 2., 5., 0.]]]], dtype=float32),
desired = array([[[[0., 0., 2., 3.],
[3., 2., 5., 0.]]]], dtype=float32),
```
This could be due to the small discrepancy in the implementations noted
[here](https://github.com/apache/tvm/pull/13163#discussion_r1002119311). I
think, the best is to track it in an issue for now.
For the record, I'm triggering the CI now to see what would happen without
the skip.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
