Add _foreach_addc(div/mul)(_).Tensor #88157
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/88157
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ✅ No FailuresAs of commit 1fedef8: This comment was automatically generated by Dr. CI and updates every 15 minutes. |
test/test_foreach.py
Outdated
| for scalar in Scalars: | ||
| self._test_pointwise_op(device, dtype, op, N, True, disable_fastpath, values=scalar) | ||
| for _, scalarlist in getScalarLists(N): | ||
| self._test_pointwise_op(device, dtype, op, N, True, disable_fastpath, values=torch.tensor(scalar)) |
There was a problem hiding this comment.
Should you also test the case where torch.tensor(scalar, device=device) ?
Same for all the updates below.
There was a problem hiding this comment.
Yes, we'll need to add more "assertRegexFail" type tests
| CUDA: foreach_tensor_addcdiv_scalarlist_cuda_ | ||
| autogen: _foreach_addcdiv.ScalarList_out | ||
|
|
||
| - func: _foreach_addcdiv_.Tensor(Tensor(a!)[] self, Tensor[] tensor1, Tensor[] tensor2, Tensor scalars) -> () |
There was a problem hiding this comment.
Curious why this is not Tensor[] scalars? That sounds more consistent?
There was a problem hiding this comment.
Because those are more difficult to convert to scalars efficiently.
There was a problem hiding this comment.
At least, #88173 will pass Tensor[] to this according to the temporary code in there.
There was a problem hiding this comment.
Well it can convert the step sizes to a torch.Tensor before calling into foreach
There was a problem hiding this comment.
Doing a .cat() before every call sounds like a big overhead?
There was a problem hiding this comment.
It's for 1-element tensors on cpu only, should be fast.
cpuhrsch
changed the title
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Support passing value scalars as a flat 1D Tensor. Currently we can only pass either an individual scalar or a ScalarList. Pull Request resolved: pytorch#88157 Approved by: https://github.com/ngimel, https://github.com/albanD
Support passing value scalars as a flat 1D Tensor. Currently we can only pass either an individual scalar or a ScalarList. Pull Request resolved: pytorch#88157 Approved by: https://github.com/ngimel, https://github.com/albanD
4 participants
Support passing value scalars as a flat 1D Tensor.
Currently we can only pass either an individual scalar or a ScalarList.