microsoft/mscclpp
1
0
Fork 0
mirror of https://github.com/microsoft/mscclpp.git synced 2026-07-17 17:27:19 +00:00
Code Issues Packages Projects Releases Wiki Activity

Enable efficient 16-rank HT EP

Browse Source 
Extend direct-fabric HT launch and runtime support to 16 ranks. Decouple TMA contributor staging from global rank discovery and restore adaptive warp geometry.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This commit is contained in:
Binyang Li
2026-07-15 06:35:34 +00:00
parent bb58d5e42b
commit 5ebbd340bb
6 changed files with 49 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
python/mscclpp/ep/README.md
View File

@@ -152,7 +152,7 @@ specialized advanced path.
The active implementation supports `mode=MoEMode.LOW_LATENCY` and
`mode=MoEMode.HIGH_THROUGHPUT`. `mode` must be a `MoEMode` enum value, not a
string. LL uses an expert-major output layout. HT uses a flat output layout and
supports only 2, 4, or 8 ranks within one detected GPU IPC/NVL fabric domain;
supports 2, 4, 8, or 16 ranks within one detected GPU IPC/NVL fabric domain;
that domain may span multiple hosts.
```python