Added symbolic shape testing for AOTAutograd#85789
Added symbolic shape testing for AOTAutograd#85789Chillee wants to merge 6 commits intogh/chillee/134/basefrom
Conversation
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/85789
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 FailuresAs of commit 88f0062: This comment was automatically generated by Dr. CI and updates every 15 minutes. |
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
ezyang
left a comment
There was a problem hiding this comment.
The use of backward() for the testing is weird but I guess it doesn't matter
| @skipIfNoSympy | ||
| @patch("functorch.compile.config.use_dynamic_shapes", True) | ||
| @patch("functorch.compile.config.use_fake_tensor", True) | ||
| @patch("functorch.compile.config.use_functionalize", False) |
There was a problem hiding this comment.
We should beef up functionalization testing too eventually
| t_args = [sample_input.input] + list(sample_input.args) | ||
| t_kwargs = sample_input.kwargs | ||
| flat_args, args_spec = pytree.tree_flatten((t_args, t_kwargs)) | ||
| sentinel_val = -42 |
There was a problem hiding this comment.
A brief explainer on this would help!
There was a problem hiding this comment.
If the sentinel does not have to be an int, allocating a fresh object and testing object identity on it is better as it is guaranteed not to collide
| if isinstance(out, tuple): | ||
| sm = 0 | ||
| for i in out: | ||
| sm += i.sum() |
There was a problem hiding this comment.
This will inplace right? Well I guess we support differentiating it so it's fine?
|
@pytorchbot revert -m "Sorry, reverting as this breaks an aot_autograd mac test on functorch. #85794 was reverted before but it was at the top of the stack so the revert still fail https://hud.pytorch.org/pytorch/pytorch/commit/823dc33b00b811c28a3924a6f0a31ba6afee7272" -c nosignal |
|
@pytorchbot successfully started a revert job. Check the current status here. |
Reverting PR 85789 failedReason: Command Details for Dev Infra teamRaised by workflow job |
Pull Request resolved: pytorch#85789 Approved by: https://github.com/ezyang
Pull Request resolved: #85789 Approved by: https://github.com/ezyang
Stack from ghstack (oldest at bottom):