Conversation
justinchuby
force-pushed
the
justinchu/ort-1.15
branch
from
f8ffece to
809b189
Compare
Signed-off-by: Justin Chu <[email protected]>
justinchuby
force-pushed
the
justinchu/ort-1.15
branch
from
a970f06 to
1d37b3b
Compare
| test_class_name="TestOutputConsistencyFullGraph", | ||
| ), | ||
| xfail( | ||
| "nn.functional.upsample_nearest2d", |
There was a problem hiding this comment.
upsample_bilinear2d is fixed but upsample_nearest2d is not. Weird
Yes it looks like they were discovered. I didn’t look too closely at what happened |
|
Strange, but won't block on this. Could you follow up? |
|
Actually wait... do we want to release with Opset19? cc @abock |
|
Sure #776 |
|
We should probably not include Opset 19 in this PR? I think while you were away @justinchuby I added support for skipping opsets in |
|
In fact, yes, let's split Opset 19 effort out of this PR since I really need ORT 1.15 for #773. |
abock
left a comment
abock
left a comment
There was a problem hiding this comment.
Re-run opgen to drop Opset 19 and open a separate PR for that.
python -m opgen -x 19 -x 20
|
Without opset19 the backend tests fail. I can dig a little deeper |
abock
left a comment
abock
left a comment
There was a problem hiding this comment.
After speaking with Justin, I had misunderstood why Opset 19 was coming in. We should take it as part of the bump.
3 participants
This change updates ORT and ort-nightly to the latest versions and cleans up the skipped tests. It also generates opset19 so that the backend tests can run.