Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Uplift third_party/tt-metal to 182c42462cd61515f60a11a46ee84f303ed073ee 2025-01-23 #1935

Merged
merged 5 commits into from
Jan 24, 2025

Conversation

vmilosevic
Copy link
Contributor

@vmilosevic vmilosevic commented Jan 23, 2025

This PR uplifts the third_party/tt-metal to the 182c42462cd61515f60a11a46ee84f303ed073ee

  • update type name (QueryResponse is now ConstraintQueryResponse)
  • Remove reference to Tensor constructor deprecated in metal commit ba35246703
  • Update json cpmcache hash include dir and use temp workaround to set -DCMAKE_CXX_SCAN_FOR_MODULES=FALSE
  • Update copy_tile_init API to accept CB argument

@@ -39,7 +39,8 @@ RUN wget https://apt.llvm.org/llvm.sh && \
./llvm.sh 17 && \
apt install -y libc++-17-dev libc++abi-17-dev && \
ln -s /usr/bin/clang-17 /usr/bin/clang && \
ln -s /usr/bin/clang++-17 /usr/bin/clang++
ln -s /usr/bin/clang++-17 /usr/bin/clang++ && \
ln -s /usr/bin/clang-scan-deps-17 /usr/bin/clang-scan-deps
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Don't love this change (is it also needed for dockers outside CI?), what is it fixing? Include path updates make sense and are typical for missing headers though, I like that part.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm getting an error CMAKE_CXX_COMPILER_CLANG_SCAN_DEPS-NOTFOUND when building locally and in CI after the metal cmake change
This does fix it locally.
I also tried adding the cmake flag directly, but that didn't work for some reason.

I assume we get this error because something upstream is using scan deps. Probably the json dep update since it's been years between the updated and prev version.

@brataTT brataTT force-pushed the uplift branch 2 times, most recently from ab65fac to 70cf4a4 Compare January 23, 2025 19:18
@@ -81,7 +81,7 @@ createOwnedTensor(std::shared_ptr<void> data,

return ::ttnn::Tensor(
createStorage<OwnedStorage>(data.get(), numElements, dataType),
::ttnn::Shape(small_vector_shape), utils::toTTNNDataType(dataType),
::ttnn::SimpleShape(small_vector_shape), utils::toTTNNDataType(dataType),
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Last month explicit ShapeBase(std::span<const uint32_t>) was added, we can now use ::ttnn::SimpleShape(shape) directly.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for noticing 🙏
I'll make the change.

@kmabeeTT
Copy link
Contributor

kmabeeTT commented Jan 24, 2025

Brings 31 tt-metal commits.

  • pass tt-mlir CI after few fixes here for issues found
  • passes tt-forge-fe CI (link)
  • passes tt-xla CI after reverting shardy change in tt-mlir on branch (it needs toolchain recompile) - link
  • passes tt-torch CI (link)
    • was failing before: (built on first attempt with shardy changes but hit runtime undefined symbol errors, Usman rebuilding toolchain job). Tried again but got ERROR: Wheel 'torch-mlir' located at /tmp/pip-unpack-d0gp8095/torch_mlir-20250123.353-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl is invalid. on job. Tried again
git l 78f2c0b774ab48bf243093d3111717a8c880dc78..182c42462cd61515f60a11a46ee84f303ed073ee
2025-01-22 182c42462c by GitHub (Author [email protected]) : Address some CMake warnings (#16993)
2025-01-22 e76afc874e by Michael Chiou (Author [email protected]) : #0: add default BH label
2025-01-22 335db8b3d4 by GitHub (Author [email protected]) : [TT-Train] Fix Taskflow test leakage  (#17004)
2025-01-22 09349dd027 by GitHub (Author [email protected]) : Replace usage of get_shape() with get_logical_shape() in more places (#16739)
2025-01-22 03260e04e8 by Michael Chiou (Author [email protected]) : #0: fix blachole scheduled post-commit
2025-01-22 818d9375e4 by Raymond Kim (Author [email protected]) : #0: [skip ci] bump functional yolov4 performance on device
2025-01-22 64d109b20f by GitHub (Author [email protected]) : #0: [skip ci] Add t3k model perf, tg demos, tgg model perf to, and remove unneeded single-card device perf from package and release (#16994)
2025-01-22 db97fb6638 by GitHub (Author [email protected]) : #0: fix small kernel Unet perf (#16987)
2025-01-22 4eeac5b86e by GitHub (Author [email protected]) : Remove mystery unused zlib dependency (#16983)
2025-01-22 02afcf39e2 by Paul Keller (Author [email protected]) : Optimize CB init for BH
2025-01-22 765c04959b by Paul Keller (Author [email protected]) : Force inline CB init
2025-01-22 a02c64b398 by Paul Keller (Author [email protected]) : Gate WATCHER_NO_INLINE w/ WATCHER_ENABLED
2025-01-22 375530a1ac by Michael Chiou (Author [email protected]) : #0: Added custom tags for BH Post commit
2025-01-22 810edfa815 by GitHub (Author [email protected]) : [skip ci] rm -rf clang-tidy-bot (#16990)
2025-01-22 09d2dc4d26 by GitHub (Author [email protected]) : Create an API for running and measuring the runtime of a ttnn op chain for use during forge compilation (#16921)
2025-01-22 ba35246703 by GitHub (Author [email protected]) : Remove deprecated Tensor constructor with Shape (#16955)
2025-01-22 de53975387 by GitHub (Author [email protected]) : Adding a toolchain file (#15581)
2025-01-22 6e2fd36789 by GitHub (Author [email protected]) : #15450: Remove default values from circular buffer parameters in LLK compute APIs: Tile Move Copy (#16664)
2025-01-22 601dcce19a by GitHub (Author [email protected]) : #16013: Increase concat heads test coverage (#16972)
2025-01-22 9262e6d40a by GitHub (Author [email protected]) : [tt-train] Add gradient norm clipping (#16771)
2025-01-22 7934fe1119 by GitHub (Author [email protected]) : [skip ci] Add initial version of CNN bring-up and optimization guide
2025-01-22 404aba8686 by GitHub (Author [email protected]) : #16557: Support padded inputs in SDPA (#16940)
2025-01-22 f246d89975 by Abhishek Agarwal (Author [email protected]) : Rebase and fix tests
2025-01-22 ada6efbc4f by Abhishek Agarwal (Author [email protected]) :     Add gatekeeper to manage sockets.     Add apis to Open/Connect datagram sockets.     Add Socket sanity test.
2025-01-22 b6957c4865 by GitHub (Author [email protected]) : #12662: pad generic reduce op input (#16925)
2025-01-22 1c5542a485 by GitHub (Author [email protected]) : #5424: Clean up Sfpu Sign kernel api (#16809)
2025-01-22 7bf259cd64 by GitHub (Author [email protected]) : #15246: Add sweeps for acos_bw, acosh_bw, atan_bw fill_zero_bw, frac_bw, log_sigmoid_bw, rad2deg_bw, trunc_bw sharded (#16372)
2025-01-22 388434991f by GitHub (Author [email protected]) : Fixing variable name to build umd tests (#16967)
2025-01-22 fe3ddefefa by Uma Devi Selvaraj (Author [email protected]) : #16510: Separate mul and silu test files
2025-01-22 29983a16e7 by Uma Devi Selvaraj (Author [email protected]) : #16510: Add sweep tests of llama add and mul op
2025-01-22 0cf1409280 by GitHub (Author [email protected]) : Add sweeps with pre-allocated output for topk and argmax (#16898)

@kmabeeTT
Copy link
Contributor

I paused tt-metal auto-uplifts so the 1am job didn't destroy this branch to let above tt-torch CI finish. Now that it finished, going to merge this and we'll kick off another auto uplift, can include the suggested Shape change on next one.

@kmabeeTT kmabeeTT merged commit 46b2891 into main Jan 24, 2025
25 checks passed
@kmabeeTT kmabeeTT deleted the uplift branch January 24, 2025 13:10
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging this pull request may close these issues.

7 participants