expose vllm.model_executor.layers.fused_moe.get_config_file_name

benchmarks/kernels/benchmark_mixtral_moe.py

@@ -4,7 +4,7 @@
 
 os.environ['CUDA_VISIBLE_DEVICES'] = '0'
 
-from vllm.model_executor.layers.fused_moe.fused_moe import fused_moe, get_config_file_name
+from vllm.model_executor.layers.fused_moe import fused_moe, get_config_file_name
 import torch
 import torch.nn.functional as F
 import triton

vllm/model_executor/layers/fused_moe/__init__.py

@@ -1,5 +1,6 @@
-from vllm.model_executor.layers.fused_moe.fused_moe import fused_moe
+from vllm.model_executor.layers.fused_moe.fused_moe import fused_moe, get_config_file_name
 
 __all__ = [
     "fused_moe",
+    "get_config_file_name",
 ]