[MoE Refactor] FusedMoE/MoERunner inversion refactor (#41184)

Signed-off-by: Bill Nell <bnell@redhat.com>
This commit is contained in:
bnellnm
2026-06-08 10:42:58 -04:00
committed by GitHub
parent 753e9d55e6
commit dc68bd8c41
90 changed files with 2738 additions and 2031 deletions
@@ -63,7 +63,6 @@ def _make_router(eplb_state: EplbLayerState | None = None) -> DummyRouter:
top_k=2,
global_num_experts=16,
eplb_state=eplb_state,
indices_type_getter=None,
)
@@ -115,6 +114,9 @@ def test_base_router_capture_with_eplb_enabled():
def test_gpu_model_runner_binds_router_capture(monkeypatch):
from vllm.v1.worker import gpu_model_runner as gmr
class _DummyRouter:
_routing_replay_out: torch.Tensor | None = None
class DummyFusedMoE:
def __init__(self):
self.layer_id = 7
@@ -132,7 +134,7 @@ def test_gpu_model_runner_binds_router_capture(monkeypatch):
# Patch the runtime import inside _bind_routed_experts_capturer.
import vllm.model_executor.layers.fused_moe.layer as fused_moe_layer
monkeypatch.setattr(fused_moe_layer, "FusedMoE", DummyFusedMoE)
monkeypatch.setattr(fused_moe_layer, "MoERunner", DummyFusedMoE)
dummy_self = types.SimpleNamespace(
compilation_config=types.SimpleNamespace(
@@ -171,7 +173,7 @@ def test_gpu_model_runner_binding_stage(monkeypatch):
import vllm.model_executor.layers.fused_moe.layer as fused_moe_layer
monkeypatch.setattr(fused_moe_layer, "FusedMoE", DummyFusedMoE)
monkeypatch.setattr(fused_moe_layer, "MoERunner", DummyFusedMoE)
dummy_self = types.SimpleNamespace(
compilation_config=types.SimpleNamespace(