K3: Move LatentMoERunner (#50678)

Signed-off-by: Jee Jee Li <jeejeelee@inferact.ai>
This commit is contained in:
Jee Jee Li
2026-08-03 09:36:17 +00:00
committed by GitHub
parent d9dac2b3d4
commit b9d1e2437e
2 changed files with 4 additions and 5 deletions
@@ -11,13 +11,12 @@ from vllm.distributed import (
tensor_model_parallel_all_reduce,
)
from vllm.logger import init_logger
from vllm.model_executor.layers.fused_moe.runner.moe_runner import MoERunner, _unpack
from vllm.model_executor.layers.layernorm import RMSNorm
from vllm.platforms import current_platform
from vllm.utils.multi_stream_utils import maybe_execute_in_parallel
from vllm.utils.torch_utils import aux_stream
from .moe_runner import MoERunner, _unpack
logger = init_logger(__name__)
+3 -3
View File
@@ -30,9 +30,6 @@ from vllm.model_executor.layers.fused_moe.router.gate_linear import GateLinear
from vllm.model_executor.layers.fused_moe.router.grouped_topk_router import (
fused_grouped_topk,
)
from vllm.model_executor.layers.fused_moe.runner.latent_moe_runner import (
LatentMoERunner,
)
from vllm.model_executor.layers.layernorm import RMSNorm
from vllm.model_executor.layers.linear import (
MergedColumnParallelLinear,
@@ -97,6 +94,9 @@ from vllm.models.common.ops.sequence_parallel import (
from vllm.models.deepseek_v4.nvidia.model import DeepseekV4MegaMoEExperts
from vllm.models.deepseek_v4.nvidia.ops.prepare_megamoe import prepare_megamoe_inputs
from vllm.models.kimi_k3.nvidia.kda import KimiK3DeltaAttention
from vllm.models.kimi_k3.nvidia.latent_moe_runner import (
LatentMoERunner,
)
from vllm.models.kimi_k3.nvidia.low_latency_gemm import (
enable_kimi_k3_low_latency_gemm,
)