From d9270b8c6ad35a29b581e40cb952e145fe9418a4 Mon Sep 17 00:00:00 2001 From: hhwxw Date: Wed, 29 Apr 2026 05:00:13 -0400 Subject: [PATCH] fix(moe): relocate orphan tuned configs after #23019 (#24004) --- benchmark/kernels/fused_moe_triton/README.md | 2 +- .../multi_node_deployment/lws_pd/lws_pd_deploy.mdx | 8 ++++---- ...=192,N=192,device_name=NVIDIA_B200,dtype=fp8_w8a8.json | 0 ...E=192,N=192,device_name=NVIDIA_H20,dtype=fp8_w8a8.json | 0 ...,N=192,device_name=NVIDIA_H20,dtype=fp8_w8a8_down.json | 0 ...92,N=192,device_name=NVIDIA_H20-3e,dtype=fp8_w8a8.json | 0 ...192,device_name=NVIDIA_H20-3e,dtype=fp8_w8a8_down.json | 0 .../E=192,N=192,device_name=NVIDIA_H20-3e.json | 0 .../E=192,N=192,device_name=NVIDIA_H20-3e_down.json | 0 .../triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20.json | 0 .../E=192,N=192,device_name=NVIDIA_H20_down.json | 0 .../triton_3_5_1/E=32,N=1792,device_name=NVIDIA_B200.json | 0 .../E=32,N=1792,device_name=NVIDIA_H100_80GB_HBM3.json | 0 .../triton_3_5_1/E=32,N=224,device_name=NVIDIA_B200.json | 0 .../E=32,N=224,device_name=NVIDIA_H100_80GB_HBM3.json | 0 .../triton_3_5_1/E=32,N=448,device_name=NVIDIA_B200.json | 0 .../E=32,N=448,device_name=NVIDIA_H100_80GB_HBM3.json | 0 .../triton_3_5_1/E=32,N=896,device_name=NVIDIA_B200.json | 0 .../E=32,N=896,device_name=NVIDIA_H100_80GB_HBM3.json | 0 .../triton_3_5_1/E=64,N=1536,device_name=NVIDIA_B200.json | 0 .../E=64,N=1536,device_name=NVIDIA_H100_80GB_HBM3.json | 0 .../triton_3_5_1/E=64,N=192,device_name=NVIDIA_B200.json | 0 .../E=64,N=192,device_name=NVIDIA_H100_80GB_HBM3.json | 0 .../triton_3_5_1/E=64,N=384,device_name=NVIDIA_B200.json | 0 .../E=64,N=384,device_name=NVIDIA_H100_80GB_HBM3.json | 0 .../triton_3_5_1/E=64,N=768,device_name=NVIDIA_B200.json | 0 .../E=64,N=768,device_name=NVIDIA_H100_80GB_HBM3.json | 0 .../E=32,N=1792,device_name=AMD_Instinct_MI325X.json | 0 .../E=32,N=224,device_name=AMD_Instinct_MI325X.json | 0 .../E=32,N=448,device_name=AMD_Instinct_MI325X.json | 0 .../E=32,N=896,device_name=AMD_Instinct_MI325X.json | 0 .../E=64,N=1536,device_name=AMD_Instinct_MI325X.json | 0 .../E=64,N=192,device_name=AMD_Instinct_MI325X.json | 0 .../E=64,N=384,device_name=AMD_Instinct_MI325X.json | 0 .../E=64,N=768,device_name=AMD_Instinct_MI325X.json | 0 35 files changed, 5 insertions(+), 5 deletions(-) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_B200,dtype=fp8_w8a8.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20,dtype=fp8_w8a8.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20,dtype=fp8_w8a8_down.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e,dtype=fp8_w8a8.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e,dtype=fp8_w8a8_down.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e_down.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20_down.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=32,N=1792,device_name=NVIDIA_B200.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=32,N=1792,device_name=NVIDIA_H100_80GB_HBM3.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=32,N=224,device_name=NVIDIA_B200.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=32,N=224,device_name=NVIDIA_H100_80GB_HBM3.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=32,N=448,device_name=NVIDIA_B200.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=32,N=448,device_name=NVIDIA_H100_80GB_HBM3.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=32,N=896,device_name=NVIDIA_B200.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=32,N=896,device_name=NVIDIA_H100_80GB_HBM3.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=64,N=1536,device_name=NVIDIA_B200.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=64,N=1536,device_name=NVIDIA_H100_80GB_HBM3.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=64,N=192,device_name=NVIDIA_B200.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=64,N=192,device_name=NVIDIA_H100_80GB_HBM3.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=64,N=384,device_name=NVIDIA_B200.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=64,N=384,device_name=NVIDIA_H100_80GB_HBM3.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=64,N=768,device_name=NVIDIA_B200.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_5_1/E=64,N=768,device_name=NVIDIA_H100_80GB_HBM3.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_6_0/E=32,N=1792,device_name=AMD_Instinct_MI325X.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_6_0/E=32,N=224,device_name=AMD_Instinct_MI325X.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_6_0/E=32,N=448,device_name=AMD_Instinct_MI325X.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_6_0/E=32,N=896,device_name=AMD_Instinct_MI325X.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_6_0/E=64,N=1536,device_name=AMD_Instinct_MI325X.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_6_0/E=64,N=192,device_name=AMD_Instinct_MI325X.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_6_0/E=64,N=384,device_name=AMD_Instinct_MI325X.json (100%) rename python/sglang/srt/layers/moe/{fused_moe_triton => moe_runner/triton_utils}/configs/triton_3_6_0/E=64,N=768,device_name=AMD_Instinct_MI325X.json (100%) diff --git a/benchmark/kernels/fused_moe_triton/README.md b/benchmark/kernels/fused_moe_triton/README.md index f11c6541a..e2bfcc41d 100644 --- a/benchmark/kernels/fused_moe_triton/README.md +++ b/benchmark/kernels/fused_moe_triton/README.md @@ -151,7 +151,7 @@ After tuning, configuration files will be generated: - **Standard tuning**: `E=64,N=640,device_name=NVIDIA_GeForce_RTX_4090,dtype=fp8_w8a8.json` - **Separate kernel tuning**: Two files for up/down kernels with TMA optimization flags -Move these files to `sglang/srt/layers/moe/fused_moe_triton/configs/triton_version/` directory to use them in SGLang. +Move these files to `sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_version/` directory to use them in SGLang. ### Supported Models diff --git a/docs_new/docs/references/multi_node_deployment/lws_pd/lws_pd_deploy.mdx b/docs_new/docs/references/multi_node_deployment/lws_pd/lws_pd_deploy.mdx index 3596d36fd..14eac03fd 100644 --- a/docs_new/docs/references/multi_node_deployment/lws_pd/lws_pd_deploy.mdx +++ b/docs_new/docs/references/multi_node_deployment/lws_pd/lws_pd_deploy.mdx @@ -147,7 +147,7 @@ spec: name: model - mountPath: /dev/infiniband name: ib - - mountPath: /sgl-workspace/sglang/python/sglang/srt/layers/moe/fused_moe_triton/configs + - mountPath: /sgl-workspace/sglang/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs name: cf - mountPath: /root/.cache name: sgl-cache @@ -305,7 +305,7 @@ spec: name: model - mountPath: /dev/infiniband name: ib - - mountPath: /sgl-workspace/sglang/python/sglang/srt/layers/moe/fused_moe_triton/configs + - mountPath: /sgl-workspace/sglang/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs name: cf dnsPolicy: ClusterFirstWithHostNet hostIPC: true @@ -458,7 +458,7 @@ spec: name: model - mountPath: /dev/infiniband name: ib - - mountPath: /sgl-workspace/sglang/python/sglang/srt/layers/moe/fused_moe_triton/configs + - mountPath: /sgl-workspace/sglang/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs name: cf dnsPolicy: ClusterFirstWithHostNet hostIPC: true @@ -592,7 +592,7 @@ spec: name: model - mountPath: /dev/infiniband name: ib - - mountPath: /sgl-workspace/sglang/python/sglang/srt/layers/moe/fused_moe_triton/configs + - mountPath: /sgl-workspace/sglang/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs name: cf dnsPolicy: ClusterFirstWithHostNet hostIPC: true diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_B200,dtype=fp8_w8a8.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_B200,dtype=fp8_w8a8.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_B200,dtype=fp8_w8a8.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_B200,dtype=fp8_w8a8.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20,dtype=fp8_w8a8.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20,dtype=fp8_w8a8.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20,dtype=fp8_w8a8.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20,dtype=fp8_w8a8.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20,dtype=fp8_w8a8_down.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20,dtype=fp8_w8a8_down.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20,dtype=fp8_w8a8_down.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20,dtype=fp8_w8a8_down.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e,dtype=fp8_w8a8.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e,dtype=fp8_w8a8.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e,dtype=fp8_w8a8.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e,dtype=fp8_w8a8.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e,dtype=fp8_w8a8_down.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e,dtype=fp8_w8a8_down.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e,dtype=fp8_w8a8_down.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e,dtype=fp8_w8a8_down.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e_down.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e_down.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e_down.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20-3e_down.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20_down.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20_down.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20_down.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=192,N=192,device_name=NVIDIA_H20_down.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=32,N=1792,device_name=NVIDIA_B200.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=32,N=1792,device_name=NVIDIA_B200.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=32,N=1792,device_name=NVIDIA_B200.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=32,N=1792,device_name=NVIDIA_B200.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=32,N=1792,device_name=NVIDIA_H100_80GB_HBM3.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=32,N=1792,device_name=NVIDIA_H100_80GB_HBM3.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=32,N=1792,device_name=NVIDIA_H100_80GB_HBM3.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=32,N=1792,device_name=NVIDIA_H100_80GB_HBM3.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=32,N=224,device_name=NVIDIA_B200.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=32,N=224,device_name=NVIDIA_B200.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=32,N=224,device_name=NVIDIA_B200.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=32,N=224,device_name=NVIDIA_B200.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=32,N=224,device_name=NVIDIA_H100_80GB_HBM3.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=32,N=224,device_name=NVIDIA_H100_80GB_HBM3.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=32,N=224,device_name=NVIDIA_H100_80GB_HBM3.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=32,N=224,device_name=NVIDIA_H100_80GB_HBM3.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=32,N=448,device_name=NVIDIA_B200.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=32,N=448,device_name=NVIDIA_B200.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=32,N=448,device_name=NVIDIA_B200.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=32,N=448,device_name=NVIDIA_B200.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=32,N=448,device_name=NVIDIA_H100_80GB_HBM3.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=32,N=448,device_name=NVIDIA_H100_80GB_HBM3.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=32,N=448,device_name=NVIDIA_H100_80GB_HBM3.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=32,N=448,device_name=NVIDIA_H100_80GB_HBM3.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=32,N=896,device_name=NVIDIA_B200.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=32,N=896,device_name=NVIDIA_B200.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=32,N=896,device_name=NVIDIA_B200.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=32,N=896,device_name=NVIDIA_B200.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=32,N=896,device_name=NVIDIA_H100_80GB_HBM3.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=32,N=896,device_name=NVIDIA_H100_80GB_HBM3.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=32,N=896,device_name=NVIDIA_H100_80GB_HBM3.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=32,N=896,device_name=NVIDIA_H100_80GB_HBM3.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=64,N=1536,device_name=NVIDIA_B200.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=64,N=1536,device_name=NVIDIA_B200.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=64,N=1536,device_name=NVIDIA_B200.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=64,N=1536,device_name=NVIDIA_B200.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=64,N=1536,device_name=NVIDIA_H100_80GB_HBM3.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=64,N=1536,device_name=NVIDIA_H100_80GB_HBM3.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=64,N=1536,device_name=NVIDIA_H100_80GB_HBM3.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=64,N=1536,device_name=NVIDIA_H100_80GB_HBM3.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=64,N=192,device_name=NVIDIA_B200.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=64,N=192,device_name=NVIDIA_B200.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=64,N=192,device_name=NVIDIA_B200.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=64,N=192,device_name=NVIDIA_B200.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=64,N=192,device_name=NVIDIA_H100_80GB_HBM3.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=64,N=192,device_name=NVIDIA_H100_80GB_HBM3.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=64,N=192,device_name=NVIDIA_H100_80GB_HBM3.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=64,N=192,device_name=NVIDIA_H100_80GB_HBM3.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=64,N=384,device_name=NVIDIA_B200.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=64,N=384,device_name=NVIDIA_B200.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=64,N=384,device_name=NVIDIA_B200.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=64,N=384,device_name=NVIDIA_B200.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=64,N=384,device_name=NVIDIA_H100_80GB_HBM3.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=64,N=384,device_name=NVIDIA_H100_80GB_HBM3.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=64,N=384,device_name=NVIDIA_H100_80GB_HBM3.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=64,N=384,device_name=NVIDIA_H100_80GB_HBM3.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=64,N=768,device_name=NVIDIA_B200.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=64,N=768,device_name=NVIDIA_B200.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=64,N=768,device_name=NVIDIA_B200.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=64,N=768,device_name=NVIDIA_B200.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=64,N=768,device_name=NVIDIA_H100_80GB_HBM3.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=64,N=768,device_name=NVIDIA_H100_80GB_HBM3.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_5_1/E=64,N=768,device_name=NVIDIA_H100_80GB_HBM3.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_5_1/E=64,N=768,device_name=NVIDIA_H100_80GB_HBM3.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_6_0/E=32,N=1792,device_name=AMD_Instinct_MI325X.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_6_0/E=32,N=1792,device_name=AMD_Instinct_MI325X.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_6_0/E=32,N=1792,device_name=AMD_Instinct_MI325X.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_6_0/E=32,N=1792,device_name=AMD_Instinct_MI325X.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_6_0/E=32,N=224,device_name=AMD_Instinct_MI325X.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_6_0/E=32,N=224,device_name=AMD_Instinct_MI325X.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_6_0/E=32,N=224,device_name=AMD_Instinct_MI325X.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_6_0/E=32,N=224,device_name=AMD_Instinct_MI325X.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_6_0/E=32,N=448,device_name=AMD_Instinct_MI325X.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_6_0/E=32,N=448,device_name=AMD_Instinct_MI325X.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_6_0/E=32,N=448,device_name=AMD_Instinct_MI325X.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_6_0/E=32,N=448,device_name=AMD_Instinct_MI325X.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_6_0/E=32,N=896,device_name=AMD_Instinct_MI325X.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_6_0/E=32,N=896,device_name=AMD_Instinct_MI325X.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_6_0/E=32,N=896,device_name=AMD_Instinct_MI325X.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_6_0/E=32,N=896,device_name=AMD_Instinct_MI325X.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_6_0/E=64,N=1536,device_name=AMD_Instinct_MI325X.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_6_0/E=64,N=1536,device_name=AMD_Instinct_MI325X.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_6_0/E=64,N=1536,device_name=AMD_Instinct_MI325X.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_6_0/E=64,N=1536,device_name=AMD_Instinct_MI325X.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_6_0/E=64,N=192,device_name=AMD_Instinct_MI325X.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_6_0/E=64,N=192,device_name=AMD_Instinct_MI325X.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_6_0/E=64,N=192,device_name=AMD_Instinct_MI325X.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_6_0/E=64,N=192,device_name=AMD_Instinct_MI325X.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_6_0/E=64,N=384,device_name=AMD_Instinct_MI325X.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_6_0/E=64,N=384,device_name=AMD_Instinct_MI325X.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_6_0/E=64,N=384,device_name=AMD_Instinct_MI325X.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_6_0/E=64,N=384,device_name=AMD_Instinct_MI325X.json diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_6_0/E=64,N=768,device_name=AMD_Instinct_MI325X.json b/python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_6_0/E=64,N=768,device_name=AMD_Instinct_MI325X.json similarity index 100% rename from python/sglang/srt/layers/moe/fused_moe_triton/configs/triton_3_6_0/E=64,N=768,device_name=AMD_Instinct_MI325X.json rename to python/sglang/srt/layers/moe/moe_runner/triton_utils/configs/triton_3_6_0/E=64,N=768,device_name=AMD_Instinct_MI325X.json