[5/N] MoE Refactor: Update MoE parallelism arguments (#8658)

This commit is contained in:
Cheng Wan
2025-08-01 01:20:03 -07:00
committed by GitHub
parent c8d3a402c1
commit 6c88f6c8d9
38 changed files with 342 additions and 299 deletions

View File

@@ -34,7 +34,8 @@ class _BaseTestDynamicEPLB(CustomTestCase):
"--dp",
"2",
"--enable-dp-attention",
"--enable-deepep-moe",
"--moe-a2a-backend",
"deepep",
"--deepep-mode",
"normal",
"--disable-cuda-graph",
@@ -96,8 +97,7 @@ class TestStaticEPLB(CustomTestCase):
trust_remote_code=True,
ep_num_redundant_experts=4,
enable_dp_attention=True,
enable_deepep_moe=True,
deepep_mode="normal",
moe_a2a_backend="deepep",
disable_cuda_graph=True,
expert_distribution_recorder_mode="stat",
tp_size=2,