A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 8, 08:45 UTC
[Bugfix] fix the import path in moe test utils.py (#34245)
Signed-off-by: Marek Michalowski <marek.michalowski@arm.com>
Marek Michalowski committed
Feb 13, 2026 at 08:13 UTC
742d214d6eeb1b0c92aabae36614be6a485fb94d
1 file changed
+5
-5
tests/kernels/moe/utils.py
+5
-5
@@ -7,11 +7,6 @@ import vllm._custom_ops as ops
7
from tests.kernels.quant_utils import per_block_cast_to_int8
8
from tests.kernels.quantization.nvfp4_utils import FLOAT4_E2M1_MAX, FLOAT8_E4M3_MAX
9
from vllm.model_executor.layers.activation import SiluAndMul
10
-from vllm.model_executor.layers.fused_moe import (
11
- TritonExperts,
12
- fused_experts,
13
- fused_topk,
14
-)
10
from vllm.model_executor.layers.fused_moe.activation import MoEActivation
11
from vllm.model_executor.layers.fused_moe.config import (
12
FusedMoEConfig,
@@ -24,10 +19,15 @@ from vllm.model_executor.layers.fused_moe.fused_batched_moe import (
19
BatchedTritonExperts,
20
NaiveBatchedExperts,
21
)
22
+from vllm.model_executor.layers.fused_moe.fused_moe import (
23
+ TritonExperts,
24
+ fused_experts,
25
+)
26
from vllm.model_executor.layers.fused_moe.modular_kernel import FusedMoEModularKernel
27
from vllm.model_executor.layers.fused_moe.prepare_finalize import (
28
MoEPrepareAndFinalizeNoEP,
29
)
30
+from vllm.model_executor.layers.fused_moe.router.fused_topk_router import fused_topk
31
from vllm.model_executor.layers.fused_moe.utils import moe_kernel_quantize_input
32
from vllm.utils.deep_gemm import per_block_cast_to_fp8
33
from vllm.utils.math_utils import round_up