Make ttnn.mean support torch.mean(tensor)
#12170
Assignees
Labels
Comments
6 tasks
bbradelTT
added a commit
that referenced
this issue
Dec 20, 2024
### Ticket Link to Github Issue #12662 #14898 #13361 #12170 ### Problem description - padding caused issues for max - keepdim=False errored out ### What's changed - remove the erroring out of keepdim=False and adjust code to handle keepdim=False properly - adding padding within min/max to ensure that it's set up properly has been pushed back to a future PR ### Checklist - [x] Post commit CI passes https://github.com/tenstorrent/tt-metal/actions/runs/12432801168 - [x] Blackhole Post commit (if applicable) https://github.com/tenstorrent/tt-metal/actions/runs/12423085751 - [x] Model regression CI testing passes (if applicable) https://github.com/tenstorrent/tt-metal/actions/runs/12423092106 same as main https://github.com/tenstorrent/tt-metal/actions/runs/12422179419/job/34683976776 - [x] Device performance regression CI testing passes (if applicable) https://github.com/tenstorrent/tt-metal/actions/runs/12423088573 - [ ] **(For models and ops writers)** Full [new models](https://github.com/tenstorrent/tt-metal/actions/workflows/full-new-models-suite.yaml) tests passes - [x] New/Existing tests provide coverage for changes
|
@jdh8 I just merged in support for keepdim=False. Could you please with latest main and let me know whether what's currently there supports your needs or which cases are not working? You may need to use Release build to avoid some asserts. |
|
Closing. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
We're trying to compile
aten.mean.default(tensor), the lowered IR oftorch.mean(tensor), in tenstorrent/pytorch2.0_ttnn#144. To support this conversion, we need any of the following:keepdim=Falseforttnn.meanThe text was updated successfully, but these errors were encountered: