microsoft/mscclpp
1
0
Fork 0
mirror of https://github.com/microsoft/mscclpp.git synced 2026-05-25 07:14:40 +00:00
Code Issues Packages Projects Releases Wiki Activity

test/ext/ep: make HT test Config env-driven

Browse Source 
Allow tuning the internode HT test cfg from the environment without
editing the source. Supported variables (all optional):

  MSCCLPP_EP_NSM          (default 152)  num channels / SMs
  MSCCLPP_EP_NVL_SEND     (default 8)
  MSCCLPP_EP_NVL_RECV     (default 256)
  MSCCLPP_EP_RDMA_SEND    (default 16)
  MSCCLPP_EP_RDMA_RECV    (default 128)

The defaults match what we use for 16-node GB200 bench runs (e.g.
NVL_RECV=512 to satisfy the HT combine assert at 16 nodes).
This commit is contained in:
Qinghua Zhou
2026-05-11 20:59:12 +00:00
parent 0f60ab5d82
commit 7b06a60786
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
test/python/ext/ep/test_internode_multirank.py
View File

@@ -121,7 +121,7 @@ def main():
# Buffer config for internode HT: needs num_rdma_bytes > 0. Size buffers
# using max(hidden, bench_hidden) so the optional bench phase fits.
cfg = ep.Config(20, 8, 256, 16, 128)
cfg = ep.Config(int(os.environ.get("MSCCLPP_EP_NSM","152")), int(os.environ.get("MSCCLPP_EP_NVL_SEND","8")), int(os.environ.get("MSCCLPP_EP_NVL_RECV","256")), int(os.environ.get("MSCCLPP_EP_RDMA_SEND","16")), int(os.environ.get("MSCCLPP_EP_RDMA_RECV","128")))
_bench_on = os.environ.get("MSCCLPP_EP_BENCH", "0") == "1"
_buf_hidden = max(hidden, int(os.environ.get("MSCCLPP_EP_BENCH_HIDDEN", "0"))) if _bench_on else hidden
num_nvl_bytes = cfg.get_nvl_buffer_size_hint(_buf_hidden * x.element_size(), num_ranks)