Vendor WeightBridge library under examples/ - #4758
Conversation
Summary: Adds a snapshot of the WeightBridge library under `examples/weightbridge/`. WeightBridge performs one-sided RDMA weight transfer from a trainer to inference workers. It builds a LoadSpec that maps source shards onto each destination worker's layout once, then reuses that plan to move only the bytes each worker needs. Its `protocol="monarch"` transport runs over Monarch, which is why the library sits alongside the other examples. 83 files: the library itself (`wbridge/`), plus its own `examples/`, `tests/`, `docs/` and `pyproject.toml`. BSD license headers were added to the 72 Python files and to `tests/run_multi_group_topology_host.sh`, matching the rest of this repository. The library carried no headers of its own. Nothing builds or imports this yet. It is source only, added so the code can be read and referenced in one place. Differential Revision: D117766258
meta-cla
Bot
added
the
CLA Signed
|
@SwiftSeal03 has exported this pull request. If you are a Meta employee, you can view the originating Diff in D117766258. |
|
Hi @SwiftSeal03! Thank you for your pull request. We require contributors to sign our Contributor License Agreement, and yours needs attention. You currently have a record in our system, but the CLA is no longer valid, and will need to be resubmitted. ProcessIn order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA. Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with If you have received this in error or have any questions, please contact us at cla@meta.com. Thanks! |
1 participant
Summary:
Adds a snapshot of the WeightBridge library under
examples/weightbridge/.WeightBridge performs one-sided RDMA weight transfer from a trainer to
inference workers. It builds a LoadSpec that maps source shards onto each
destination worker's layout once, then reuses that plan to move only the bytes
each worker needs. Its
protocol="monarch"transport runs over Monarch, whichis why the library sits alongside the other examples.
83 files: the library itself (
wbridge/), plus its ownexamples/,tests/,docs/andpyproject.toml.BSD license headers were added to the 72 Python files and to
tests/run_multi_group_topology_host.sh, matching the rest of this repository.The library carried no headers of its own.
Nothing builds or imports this yet. It is source only, added so the code can be
read and referenced in one place.
Differential Revision: D117766258