EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

fix(deepep): resolve benchmark failure on 4×IB-card setup by aligning tuning config with DeepEP commit bdd119f8 (#11965)

Browse Source
This commit is contained in:
Zhengyi Lai
2025-10-23 12:20:54 +08:00
committed by GitHub
parent 007b849b0e
commit 81fd2b0ee0
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
benchmark/kernels/deepep/tuning_deepep.py
View File

@@ -381,8 +381,8 @@ def test_main(
# Tune combine performance # Tune combine performance
best_time, best_results = 1e10, None best_time, best_results = 1e10, None
for nvl_chunk_size in range(1, 5, 1): for nvl_chunk_size in range(1, 8, 1):
for rdma_chunk_size in range(8, 33, 4): for rdma_chunk_size in range(12 if num_nodes == 2 else 8, 33, 4):
config_kwargs = { config_kwargs = {
"num_sms": num_sms, "num_sms": num_sms,
"num_max_nvl_chunked_send_tokens": nvl_chunk_size, "num_max_nvl_chunked_send_tokens": nvl_chunk_size,