Jeffrey Wan

Tracker: Slow gradcheck failures possibly indicating incorrect gradients

2

### 🐛 Describe the bug For a very long time, most tests are being [skipped](https://ossci-raw-job-status.s3.amazonaws.com/log/7032407678) by slow gradcheck (see https://github.com/pytorch/pytorch/issues/80314#issuecomment-1168074007): At least the following are failing in slow gradcheck (There...

Add decomposition tensor

2

Exploring some possible UX for using decompositions with subclassing cc @ezyang @albanD

Avoid running TOR003 by default

3

Today, use_reentrant defaults to True, but TOR003 sets use_reentrant=False which may subtly differ in behavior in certain cases. We should either make TOR003 set use_reentrant=True, or not run it by...

[do not review] copy paste of custom attribute changes for testing

2

Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * #121445 * __->__ #121742 * #121373 From https://github.com/pytorch/pytorch/pull/121693 cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @chenyang78 @kadeng @chauhang

Switch to using Python nested int

1

Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * __->__ #141166 Doesn't seem to noticeably slow down eager - TestNestedTensorSubclass tests with and without the PR finished in similar amounts of time...