Dev/update readme pip install#770
Merged
Merged
Conversation
#757) ## Summary - Expose each pip-installable third_party submodule as a pip extra pinned to its exact commit: pip install "primus[megatron]" --extra-index-url <pages> (also torchtitan / maxtext / emerging-optimizers / megatron-bridge). tools/pip_index/gen_thirdparty_extras.py reads submodule pins from the current checkout and injects optional-dependencies in CI before build, so each branch/release advertises its own commits. - megatron-bridge nests Megatron-LM; pip does not fetch git submodules, so its extra also pulls primus[megatron] (megatron-core). - HummingbirdXT is excluded: no pyproject/setup.py, not pip-installable. - pyproject: enable [tool.hatch.metadata] allow-direct-references for the git+ deps. - Release timing fix: before triggering the dashboard rebuild, poll the Releases API until the uploaded wheel is visible, so the GitHub Pages pip index is no longer generated empty. ## Test plan - [x] gen_thirdparty_extras.py --dry-run lists 5 extras with correct pinned commits - [x] Injected pyproject + python -m build succeeds; wheel METADATA has Provides-Extra + git+...@commit for all 5; megatron-bridge also pulls megatron-core - [ ] On next release: pip index is populated (timing fix) and pip install "primus[megatron]" resolves the pinned commit --------- Co-authored-by: HuangWei-95 <weihuan@amd.com> Co-authored-by: Cursor <cursoragent@cursor.com>
Reorganize the Setup & Deployment section around the AMD published training Docker images and split the walkthrough into two options. - Quick Start Option 1: git clone the repo and run training in a container (now recommended), with release-branch checkout and submodule init steps - Quick Start Option 2: install the Primus wheel into a venv and run training in a container (or directly on a host) - add the MaxText (rocm/jax-training) image alongside rocm/primus - remove the older git+pip / deps-sync install walkthrough
Provide documentation and tooling to reproduce the training Docker
image's software stack directly on a host machine, for users who
cannot run containers.
- docs/install-on-host.md: step-by-step host install derived from the
training Dockerfile (ROCm via pip rocm-sdk-devel so no system ROCm or
sudo, source-built kernel libs, optional multi-node UCX/OpenMPI/
rocSHMEM, verification, and operational notes)
- tools/installation/{setup.sh,env.sh,README.md}: staged, re-runnable
venv installer with automatic GPU arch detection (gfx942/gfx950),
fail-fast error handling, and host-specific fixes (mamba via pip,
apache-tvm-ffi/z3-solver pins, Megatron helpers_cpp build)
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
GeneDer
requested review from
Xiaoming-AMD,
limou102 and
wenxie-amd
as code owners
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
3 participants
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.
No description provided.