Repository navigation
torchmonarch: Add version 0.7.0 - #3182
Merged
Merged
Conversation
Build Meta's Monarch (setuptools-rust PyO3 extension over an ~850-crate Rust workspace, plus two libtorch C++ extensions) with its CPU-only tensor engine against the registry's torch 2.14.0, mirroring upstream's release builder and its macOS CPU leg. The dashboard frontend is bundled on an x86 runner, as upstream's ci.yml does, since node.js has no riscv64 release.
luhenry
added a commit
that referenced
this pull request
Oct 14, 2026
…ning (PR #3183), auto-gptq -> parked (CUDA-only extensions; CPU build is py3-none-any, upstream unmaintained)
luhenry
marked this pull request as draft
October 14, 2026 11:31
All three interpreters failed identically after ~3h20m of compiling, deep in the ~850-crate graph: libsqlite3-sys's build script uses bindgen, which panics with "Unable to find libclang" because clang-devel (the package providing libclang.so) is not installed. The existing comment assumed clang-devel only served the CUDA/RDMA crates a CPU build skips; libsqlite3-sys needs it too.
pyarrow>=17 is a runtime dependency of the built wheel. At test-install time, pip resolves it against both pypi.org and pypi.riseproject.dev; both list 26.0.0, but only ours is a riscv64 wheel. Without a binary preference pip source-builds PyPI's sdist instead, and that needs the real Apache Arrow C++ (no FindArrow.cmake in the container), so cp313 and cp314 died in pyarrow's CMake configure. See gotcha 354.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
torchmonarch0.7.0Compiles Monarch's PyO3 extension (an ~850-crate Rust workspace) and its two libtorch C++ extensions, with the CPU-only tensor engine. Upstream publishes no riscv64 wheel.
Mirrors upstream's
publish_release.yml(build-dist.yml, and the CPU-onlybuild-macos.ymlleg).Differs from upstream
torch==2.14.0(upstream 2.14.1) - newest riscv64 torch we host.ci.ymldoes - node.js has no riscv64 release.protobuf-compiler- clang-devel/libunwind serve the CUDA/RDMA crates only.CARGO_BUILD_JOBS=2plus 16 GB swap - memory guard for the ~280 MB extension link.Matrix: cp312/cp313/cp314 - upstream ships no free-threaded build.
Testing
import monarchcheck, plus its CPU-runnabletest_extension_linking.py,test_rust_bindings_load.py,test_grad_generator.pyandtest_python_actors.py::test_stream.License: OK
Counts pending the first riscv64 CI run.