Remove moe_dense_tp_size for TP configs

recipes/gb200-fp4/1k1k/1p2d-mtp.yaml

@@ -74,7 +74,6 @@ backend:
       max-running-requests: 512
       scheduler-recv-interval: 10
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       load-balance-method: "round_robin"
       disaggregation-bootstrap-port: 30001
       data-parallel-size: 1
@@ -104,7 +103,6 @@ backend:
       cuda-graph-max-bs: 256
       scheduler-recv-interval: 10
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       tensor-parallel-size: 4
       expert-parallel-size: 1
       speculative-algorithm: "EAGLE"

recipes/gb200-fp4/1k1k/low-latency.yaml

@@ -76,7 +76,6 @@ backend:
       max-running-requests: 512
       scheduler-recv-interval: 10
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       load-balance-method: "round_robin"
       disaggregation-bootstrap-port: 30001
       disaggregation-transfer-backend: nixl
@@ -104,7 +103,6 @@ backend:
       cuda-graph-max-bs: 256
       scheduler-recv-interval: 10
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       disaggregation-transfer-backend: nixl
       fp4-gemm-backend: "flashinfer_trtllm"
       tensor-parallel-size: 4

recipes/gb200-fp4/1k8k/low-latency.yaml

@@ -74,7 +74,6 @@ backend:
       disable-cuda-graph: true
       max-running-requests: 512
       scheduler-recv-interval: 10
-      moe-dense-tp-size: 1
       load-balance-method: "round_robin"
       disaggregation-bootstrap-port: 30001
       data-parallel-size: 1
@@ -101,7 +100,6 @@ backend:
       chunked-prefill-size: 8192
       cuda-graph-max-bs: 256
       scheduler-recv-interval: 10
-      moe-dense-tp-size: 1
       tensor-parallel-size: 4
       expert-parallel-size: 1
       fp4-gemm-backend: "flashinfer_trtllm"

recipes/gb200-fp4/8k1k/low-latency.yaml

@@ -76,7 +76,6 @@ backend:
       max-running-requests: 512
       scheduler-recv-interval: 10
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       load-balance-method: "round_robin"
       disaggregation-bootstrap-port: 30001
       data-parallel-size: 1
@@ -105,7 +104,6 @@ backend:
       cuda-graph-max-bs: 256
       scheduler-recv-interval: 10
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       disaggregation-transfer-backend: nixl
       fp4-gemm-backend: "flashinfer_trtllm"
       tensor-parallel-size: 4

recipes/gb200-fp8/1k1k/low-latency-mtp.yaml

@@ -82,7 +82,6 @@ backend:
       scheduler-recv-interval: 10
       enable-flashinfer-allreduce-fusion: true
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       tensor-parallel-size: 4
       data-parallel-size: 1
       expert-parallel-size: 1
@@ -111,7 +110,6 @@ backend:
       max-running-requests: 512
       scheduler-recv-interval: 10
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       prefill-round-robin-balance: true
       tensor-parallel-size: 8
       data-parallel-size: 1

recipes/gb200-fp8/1k1k/low-latency.yaml

@@ -81,7 +81,6 @@ backend:
       scheduler-recv-interval: 10
       fp8-gemm-backend: "flashinfer_trtllm"
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       tensor-parallel-size: 4
       data-parallel-size: 1
       expert-parallel-size: 1
@@ -106,7 +105,6 @@ backend:
       max-running-requests: 128
       scheduler-recv-interval: 10
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       prefill-round-robin-balance: true
       tensor-parallel-size: 4
       data-parallel-size: 1

recipes/gb200-fp8/1k8k/low-latency-mtp.yaml

@@ -79,7 +79,6 @@ backend:
       load-balance-method: "round_robin"
       scheduler-recv-interval: 10
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       disaggregation-bootstrap-port: 30001  
       tensor-parallel-size: 8
       data-parallel-size: 1
@@ -109,7 +108,6 @@ backend:
       scheduler-recv-interval: 10
       enable-flashinfer-allreduce-fusion: false
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       prefill-round-robin-balance: true
       disaggregation-bootstrap-port: 30001  
       tensor-parallel-size: 8

recipes/gb200-fp8/1k8k/low_latency.yaml

@@ -81,7 +81,6 @@ backend:
       enable-flashinfer-allreduce-fusion: false
       fp8-gemm-backend: "flashinfer_trtllm"
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       disaggregation-bootstrap-port: 30001
       tensor-parallel-size: 8
       data-parallel-size: 1
@@ -107,7 +106,6 @@ backend:
       scheduler-recv-interval: 10
       enable-flashinfer-allreduce-fusion: false
       enable-symm-mem: false #true
-      moe-dense-tp-size: 1
       disaggregation-bootstrap-port: 30001
       prefill-round-robin-balance: true
       tensor-parallel-size: 8

recipes/gb200-fp8/8k1k/low-latency-mtp.yaml

@@ -79,7 +79,6 @@ backend:
       load-balance-method: "round_robin"
       scheduler-recv-interval: 10
       enable-flashinfer-allreduce-fusion: false
-      moe-dense-tp-size: 1
       tensor-parallel-size: 4
       data-parallel-size: 1
       expert-parallel-size: 1
@@ -107,7 +106,6 @@ backend:
       scheduler-recv-interval: 10
       enable-flashinfer-allreduce-fusion: false
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       prefill-round-robin-balance: true
       tensor-parallel-size: 8
       data-parallel-size: 1

recipes/gb200-fp8/8k1k/low-latency.yaml

@@ -77,7 +77,6 @@ backend:
       max-running-requests: 512
       load-balance-method: "round_robin"
       scheduler-recv-interval: 10
-      moe-dense-tp-size: 1
       tensor-parallel-size: 8
       data-parallel-size: 1
       expert-parallel-size: 1
@@ -102,7 +101,6 @@ backend:
       max-running-requests: 512
       scheduler-recv-interval: 10
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       prefill-round-robin-balance: true
       tensor-parallel-size: 8
       data-parallel-size: 1

recipes/gb300-fp4/128k8k/5-low-latency.yaml

@@ -73,7 +73,6 @@ backend:
       expert-parallel-size: 4
       kv-cache-dtype: fp8_e4m3
       mem-fraction-static: 0.95
-      moe-dense-tp-size: 1
       moe-runner-backend: flashinfer_trtllm
       pipeline-parallel-size: 1
       prefill-round-robin-balance: true
@@ -100,7 +99,6 @@ backend:
       load-balance-method: round_robin
       max-total-tokens: 544000
       mem-fraction-static: 0.95
-      moe-dense-tp-size: 1
       moe-runner-backend: flashinfer_trtllm
       pipeline-parallel-size: 4
       quantization: modelopt_fp4
@@ -116,4 +114,4 @@ benchmark:
   isl: 128000
   osl: 8000
   req_rate: "inf"
-  type: "sa-bench"
+  type: "sa-bench"

recipes/gb300-fp4/1k1k/low_latency.yaml

@@ -76,7 +76,6 @@ backend:
       max-running-requests: 512
       scheduler-recv-interval: 10
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       load-balance-method: "round_robin"
       disaggregation-bootstrap-port: 30001
       data-parallel-size: 1
@@ -104,7 +103,6 @@ backend:
       cuda-graph-max-bs: 256
       scheduler-recv-interval: 10
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       tensor-parallel-size: 4
       expert-parallel-size: 1
       fp4-gemm-backend: "flashinfer_trtllm"

recipes/gb300-fp4/1k8k/low-latency.yaml

@@ -74,7 +74,6 @@ backend:
       disable-cuda-graph: true
       max-running-requests: 512
       scheduler-recv-interval: 10
-      moe-dense-tp-size: 1
       load-balance-method: "round_robin"
       disaggregation-bootstrap-port: 30001
       data-parallel-size: 1
@@ -101,7 +100,6 @@ backend:
       chunked-prefill-size: 8192
       cuda-graph-max-bs: 256
       scheduler-recv-interval: 10
-      moe-dense-tp-size: 1
       tensor-parallel-size: 4
       expert-parallel-size: 1
       fp4-gemm-backend: "flashinfer_trtllm"

recipes/gb300-fp4/8k1k/low_latency.yaml

@@ -76,7 +76,6 @@ backend:
       max-running-requests: 512
       scheduler-recv-interval: 10
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       load-balance-method: "round_robin"
       disaggregation-bootstrap-port: 30001
       data-parallel-size: 1
@@ -106,7 +105,6 @@ backend:
       cuda-graph-max-bs: 128
       scheduler-recv-interval: 10
       enable-symm-mem: true
-      moe-dense-tp-size: 1
       tensor-parallel-size: 4
       expert-parallel-size: 1
       enable-dp-attention: false

recipes/gb300-fp8/1k1k/mtp/low-latency.yaml

@@ -87,7 +87,6 @@ backend:
       scheduler-recv-interval: 10
       enable-flashinfer-allreduce-fusion: false # to save mem
       enable-symm-mem: false # to save mem 
-      moe-dense-tp-size: 1
       tensor-parallel-size: 4
       data-parallel-size: 1
       expert-parallel-size: 1
@@ -117,7 +116,6 @@ backend:
       scheduler-recv-interval: 1  # save mem
       enable-flashinfer-allreduce-fusion: false # to save mem 
       enable-symm-mem: false # to save mem
-      moe-dense-tp-size: 1
       prefill-round-robin-balance: true
       tensor-parallel-size: 4
       data-parallel-size: 1

recipes/gb300-fp8/1k1k/stp/low-latency.yaml

@@ -83,7 +83,6 @@ backend:
       scheduler-recv-interval: 10
       enable-flashinfer-allreduce-fusion: false # to save mem
       enable-symm-mem: false # to save mem 
-      moe-dense-tp-size: 1
       tensor-parallel-size: 4
       data-parallel-size: 1
       expert-parallel-size: 1
@@ -109,7 +108,6 @@ backend:
       scheduler-recv-interval: 1  # save mem
       enable-flashinfer-allreduce-fusion: false # to save mem 
       enable-symm-mem: false # to save mem
-      moe-dense-tp-size: 1
       prefill-round-robin-balance: true
       tensor-parallel-size: 4
       data-parallel-size: 1

recipes/gb300-fp8/1k8k/mtp/low-latency.yaml

@@ -87,7 +87,6 @@ backend:
       scheduler-recv-interval: 10
       enable-flashinfer-allreduce-fusion: false # to save mem
       enable-symm-mem: false # to save mem 
-      moe-dense-tp-size: 1
       tensor-parallel-size: 4
       data-parallel-size: 1
       expert-parallel-size: 1
@@ -117,7 +116,6 @@ backend:
       scheduler-recv-interval: 1  # save mem
       enable-flashinfer-allreduce-fusion: false # to save mem 
       enable-symm-mem: false # to save mem
-      moe-dense-tp-size: 1
       prefill-round-robin-balance: true
       tensor-parallel-size: 4
       data-parallel-size: 1

recipes/gb300-fp8/1k8k/stp/low-latency.yaml

@@ -83,7 +83,6 @@ backend:
       scheduler-recv-interval: 10
       enable-flashinfer-allreduce-fusion: false # to save mem
       enable-symm-mem: false # to save mem 
-      moe-dense-tp-size: 1
       tensor-parallel-size: 4
       data-parallel-size: 1
       expert-parallel-size: 1
@@ -109,7 +108,6 @@ backend:
       scheduler-recv-interval: 1  # save mem
       enable-flashinfer-allreduce-fusion: false # to save mem 
       enable-symm-mem: false # to save mem
-      moe-dense-tp-size: 1
       prefill-round-robin-balance: true
       tensor-parallel-size: 4
       data-parallel-size: 1

recipes/gb300-fp8/8k1k/mtp/low-latency.yaml

@@ -87,7 +87,6 @@ backend:
       scheduler-recv-interval: 10
       enable-flashinfer-allreduce-fusion: false # to save mem
       enable-symm-mem: false # to save mem 
-      moe-dense-tp-size: 1
       tensor-parallel-size: 4
       data-parallel-size: 1
       expert-parallel-size: 1
@@ -117,7 +116,6 @@ backend:
       scheduler-recv-interval: 1  # save mem
       enable-flashinfer-allreduce-fusion: false # to save mem 
       enable-symm-mem: false # to save mem
-      moe-dense-tp-size: 1
       prefill-round-robin-balance: true
       tensor-parallel-size: 4
       data-parallel-size: 1

recipes/gb300-fp8/8k1k/stp/low-latency.yaml

@@ -83,7 +83,6 @@ backend:
       scheduler-recv-interval: 10
       enable-flashinfer-allreduce-fusion: false # to save mem
       enable-symm-mem: false # to save mem 
-      moe-dense-tp-size: 1
       tensor-parallel-size: 4
       data-parallel-size: 1
       expert-parallel-size: 1
@@ -109,7 +108,6 @@ backend:
       scheduler-recv-interval: 1  # save mem
       enable-flashinfer-allreduce-fusion: false # to save mem 
       enable-symm-mem: false # to save mem
-      moe-dense-tp-size: 1
       prefill-round-robin-balance: true
       tensor-parallel-size: 4
       data-parallel-size: 1