minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[Spec] Remove dead padded_static_len and stale SGLANG_ENABLE_SPEC_V2 references (#30998)

Browse Source
This commit is contained in:
Liangsheng Yin
2026-07-13 15:30:31 -05:00
committed by GitHub
parent 47030b28be
commit e2728ac504
65 changed files with 12 additions and 166 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs_new/docs/advanced_features/pd_disaggregation.mdx
-3
View File
@@ -493,7 +493,6 @@ export DEEPEP_NORMAL_LONG_SEQ_PER_ROUND_TOKENS=3584
export ASCEND_MF_STORE_URL="tcp://127.0.0.1:24669"
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=3600
export SGLANG_DISAGGREGATION_WAITING_TIMEOUT=3600
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=0
export DEEP_NORMAL_MODE_USE_INT8_QUANT=1
export SGLANG_DEEPEP_BF16_DISPATCH=0
@@ -553,7 +552,6 @@ export DEEPEP_NORMAL_LONG_SEQ_PER_ROUND_TOKENS=3584
export ASCEND_MF_STORE_URL="tcp://127.0.0.1:24669"
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=3600
export SGLANG_DISAGGREGATION_WAITING_TIMEOUT=3600
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=0
export DEEP_NORMAL_MODE_USE_INT8_QUANT=1
export SGLANG_DEEPEP_BF16_DISPATCH=0
@@ -678,7 +676,6 @@ do
then
echo "${D_IP[$i]}"
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export HCCL_BUFFSIZE=900
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=112
export TASK_QUEUE_ENABLE=1
docs_new/docs/hardware-platforms/ascend-npus/best_practice/deepseek_r1.mdx
-7
View File
@@ -151,7 +151,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=78
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export TASK_QUEUE_ENABLE=1
python3 -m sglang.launch_server \
@@ -279,7 +278,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=56
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_NPU_USE_MLAPO=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=200
export SGLANG_SCHEDULER_DECREASE_PREFILL_IDLE=1
@@ -457,7 +455,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=12
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SCHEDULER_SKIP_ALL_GATHER=1
export TASK_QUEUE_ENABLE=1
@@ -666,7 +663,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=64
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_LM_HEAD_TP=8
export SGLANG_NPU_FUSED_MOE_MODE=1
export TASK_QUEUE_ENABLE=1
@@ -871,7 +867,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=12
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SCHEDULER_SKIP_ALL_GATHER=1
export TASK_QUEUE_ENABLE=1
@@ -1076,7 +1071,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=12
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SCHEDULER_SKIP_ALL_GATHER=1
export TASK_QUEUE_ENABLE=1
@@ -1280,7 +1274,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=16
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SCHEDULER_SKIP_ALL_GATHER=1
export TASK_QUEUE_ENABLE=1
docs_new/docs/hardware-platforms/ascend-npus/best_practice/deepseek_v3_2.mdx
-3
View File
@@ -148,7 +148,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=8
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SCHEDULER_SKIP_ALL_GATHER=1
export TASK_QUEUE_ENABLE=0
@@ -347,7 +346,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=8
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SCHEDULER_SKIP_ALL_GATHER=1
export TASK_QUEUE_ENABLE=0
@@ -546,7 +544,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=8
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SCHEDULER_SKIP_ALL_GATHER=1
export TASK_QUEUE_ENABLE=0
docs_new/docs/hardware-platforms/ascend-npus/best_practice/glm5_1.mdx
-9
View File
@@ -84,7 +84,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=32
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -280,7 +279,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=16
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SPEC_ENABLE_OVERLAP_REFLOW=1
export TASK_QUEUE_ENABLE=0
@@ -481,7 +479,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=64
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SPEC_ENABLE_OVERLAP_REFLOW=1
export TASK_QUEUE_ENABLE=0
@@ -690,7 +687,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=16
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SPEC_ENABLE_OVERLAP_REFLOW=1
export TASK_QUEUE_ENABLE=0
@@ -893,7 +889,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=16
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SPEC_ENABLE_OVERLAP_REFLOW=1
export TASK_QUEUE_ENABLE=0
@@ -1094,7 +1089,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=64
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SPEC_ENABLE_OVERLAP_REFLOW=1
export TASK_QUEUE_ENABLE=0
@@ -1295,7 +1289,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=48
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SCHEDULER_SKIP_ALL_GATHER=1
export SGLANG_SPEC_ENABLE_OVERLAP_REFLOW=1
export TASK_QUEUE_ENABLE=0
@@ -1505,7 +1498,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=24
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SPEC_ENABLE_OVERLAP_REFLOW=1
export TASK_QUEUE_ENABLE=0
@@ -1718,7 +1710,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=40
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SPEC_ENABLE_OVERLAP_REFLOW=1
export TASK_QUEUE_ENABLE=0
docs_new/docs/hardware-platforms/ascend-npus/best_practice/kimi_k2_6.mdx
-9
View File
@@ -87,7 +87,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=64
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -278,7 +277,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=64
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_NPU_USE_MLAPO=1
export SGLANG_NPU_USE_MULTI_STREAM=1
@@ -478,7 +476,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=64
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_NPU_USE_MLAPO=1
export SGLANG_NPU_USE_MULTI_STREAM=1
@@ -678,7 +675,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=64
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_NPU_USE_MLAPO=1
export SGLANG_NPU_USE_MULTI_STREAM=1
@@ -873,7 +869,6 @@ do
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=64
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_NPU_USE_MLAPO=1
export SGLANG_NPU_USE_MULTI_STREAM=1
@@ -1009,7 +1004,6 @@ export HCCL_OP_EXPANSION_MODE=AIV
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=112
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_NPU_USE_MULTI_STREAM=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -1121,7 +1115,6 @@ export HCCL_OP_EXPANSION_MODE=AIV
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=32
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_NPU_USE_MULTI_STREAM=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -1237,7 +1230,6 @@ export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=96
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_NPU_USE_MLAPO=1
export SGLANG_NPU_USE_MULTI_STREAM=1
export SGLANG_SET_CPU_AFFINITY=1
@@ -1350,7 +1342,6 @@ export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=96
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
docs_new/docs/hardware-platforms/ascend-npus/best_practice/mimo_v2_flash.mdx
-4
View File
@@ -83,7 +83,6 @@ export SGLANG_DEEPEP_BF16_DISPATCH=0
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=3600
export SGLANG_DISAGGREGATION_WAITING_TIMEOUT=3600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=0
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -269,7 +268,6 @@ export SGLANG_DEEPEP_BF16_DISPATCH=0
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=3600
export SGLANG_DISAGGREGATION_WAITING_TIMEOUT=3600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=0
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -455,7 +453,6 @@ export SGLANG_DEEPEP_BF16_DISPATCH=0
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=3600
export SGLANG_DISAGGREGATION_WAITING_TIMEOUT=3600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=0
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -641,7 +638,6 @@ export SGLANG_DEEPEP_BF16_DISPATCH=0
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=3600
export SGLANG_DISAGGREGATION_WAITING_TIMEOUT=3600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=0
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
docs_new/docs/hardware-platforms/ascend-npus/best_practice/minimax_m2_5.mdx
-5
View File
@@ -85,7 +85,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=640
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_ENABLE_TP_MEMORY_INBALANCE_CHECK=0
export SGLANG_EXTERNAL_MODEL_PACKAGE=custom_eagle3
export SGLANG_SET_CPU_AFFINITY=1
@@ -192,7 +191,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=140000
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_EXTERNAL_MODEL_PACKAGE=custom_eagle3
export SGLANG_NPU_FUSED_MOE_MODE=2
export SGLANG_SET_CPU_AFFINITY=1
@@ -301,7 +299,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=160000
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_EXTERNAL_MODEL_PACKAGE=custom_eagle3
export SGLANG_NPU_FUSED_MOE_MODE=2
export SGLANG_SET_CPU_AFFINITY=1
@@ -411,7 +408,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=204800
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_EXTERNAL_MODEL_PACKAGE=custom_eagle3
export SGLANG_NPU_FUSED_MOE_MODE=2
export SGLANG_SET_CPU_AFFINITY=1
@@ -516,7 +512,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=204800
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_EXTERNAL_MODEL_PACKAGE=custom_eagle3
export SGLANG_NPU_FUSED_MOE_MODE=2
export SGLANG_SET_CPU_AFFINITY=1
docs_new/docs/hardware-platforms/ascend-npus/best_practice/qwen3-8b.mdx
-3
View File
@@ -77,7 +77,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=50
export SGLANG_SCHEDULER_DECREASE_PREFILL_IDLE=1
@@ -175,7 +174,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
python3 -m sglang.launch_server \
--model-path $MODEL_PATH \
@@ -271,7 +269,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
python3 -m sglang.launch_server \
--model-path $MODEL_PATH \
docs_new/docs/hardware-platforms/ascend-npus/best_practice/qwen3_235b_a22b.mdx
-2
View File
@@ -77,7 +77,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
python3 -m sglang.launch_server \
--model-path $MODEL_PATH \
@@ -175,7 +174,6 @@ export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=188416
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_NPU_FUSED_MOE_MODE=2
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=100
export SGLANG_SCHEDULER_DECREASE_PREFILL_IDLE=1
docs_new/docs/hardware-platforms/ascend-npus/best_practice/qwen3_30b_a3b.mdx
-4
View File
@@ -81,7 +81,6 @@ export INF_NAN_MODE_FORCE_DISABLE=1
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:False
export SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN=1
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=200
export SGLANG_SCHEDULER_DECREASE_PREFILL_IDLE=1
export SGLANG_USE_MAX_DP_ATT=1
@@ -184,7 +183,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=200
export SGLANG_SCHEDULER_DECREASE_PREFILL_IDLE=1
@@ -284,7 +282,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=200
export SGLANG_SCHEDULER_DECREASE_PREFILL_IDLE=1
@@ -383,7 +380,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SET_CPU_AFFINITY=1
export TRANSFORMERS_VERBOSITY=error
docs_new/docs/hardware-platforms/ascend-npus/best_practice/qwen3_32b.mdx
-3
View File
@@ -77,7 +77,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=200
export SGLANG_SCHEDULER_DECREASE_PREFILL_IDLE=1
@@ -174,7 +173,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_NPU_USE_DEEPGEMM=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=100
export SGLANG_SCHEDULER_DECREASE_PREFILL_IDLE=1
@@ -273,7 +271,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=600
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_NPU_USE_DEEPGEMM=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=100
export SGLANG_SCHEDULER_DECREASE_PREFILL_IDLE=1
docs_new/docs/hardware-platforms/ascend-npus/best_practice/qwen3_5_397b.mdx
-10
View File
@@ -87,7 +87,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=128
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_ENABLE_TP_MEMORY_INBALANCE_CHECK=0
export SGLANG_SET_CPU_AFFINITY=1
export SGLANG_ZBAL_BOOTSTRAP_URL=tcp://127.0.0.1:24669
@@ -198,7 +197,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=128
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_ENABLE_TP_MEMORY_INBALANCE_CHECK=0
export SGLANG_SET_CPU_AFFINITY=1
export SGLANG_ZBAL_BOOTSTRAP_URL=tcp://127.0.0.1:24669
@@ -309,7 +307,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=128
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -419,7 +416,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=128
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_ENABLE_TP_MEMORY_INBALANCE_CHECK=0
export SGLANG_SET_CPU_AFFINITY=1
export SGLANG_ZBAL_BOOTSTRAP_URL=tcp://127.0.0.1:24669
@@ -534,7 +530,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=128
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_ENABLE_TP_MEMORY_INBALANCE_CHECK=0
export SGLANG_SET_CPU_AFFINITY=1
export SGLANG_ZBAL_BOOTSTRAP_URL=tcp://127.0.0.1:24669
@@ -649,7 +644,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=128
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_ENABLE_TP_MEMORY_INBALANCE_CHECK=0
export SGLANG_SET_CPU_AFFINITY=1
export SGLANG_ZBAL_BOOTSTRAP_URL=tcp://127.0.0.1:24669
@@ -763,7 +757,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=128
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_ENABLE_TP_MEMORY_INBALANCE_CHECK=0
export SGLANG_SET_CPU_AFFINITY=1
export SGLANG_ZBAL_BOOTSTRAP_URL=tcp://127.0.0.1:24669
@@ -877,7 +870,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=128
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_ENABLE_TP_MEMORY_INBALANCE_CHECK=0
export SGLANG_SET_CPU_AFFINITY=1
export SGLANG_ZBAL_BOOTSTRAP_URL=tcp://127.0.0.1:24669
@@ -992,7 +984,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=128
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_ENABLE_TP_MEMORY_INBALANCE_CHECK=0
export SGLANG_SET_CPU_AFFINITY=1
export SGLANG_ZBAL_BOOTSTRAP_URL=tcp://127.0.0.1:24669
@@ -1107,7 +1098,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=128
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
docs_new/docs/hardware-platforms/ascend-npus/best_practice/qwen3_6_27b.mdx
-7
View File
@@ -75,7 +75,6 @@ export HCCL_OP_EXPANSION_MODE=AIV
export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=150
export SGLANG_SCHEDULER_DECREASE_PREFILL_IDLE=1
export SGLANG_SET_CPU_AFFINITY=1
@@ -174,7 +173,6 @@ export HCCL_OP_EXPANSION_MODE=AIV
export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=150
export SGLANG_SCHEDULER_DECREASE_PREFILL_IDLE=1
export SGLANG_SET_CPU_AFFINITY=1
@@ -274,7 +272,6 @@ export HCCL_OP_EXPANSION_MODE=AIV
export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -376,7 +373,6 @@ export HCCL_OP_EXPANSION_MODE=AIV
export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=0
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=130
export SGLANG_SCHEDULER_DECREASE_PREFILL_IDLE=1
export SGLANG_SET_CPU_AFFINITY=1
@@ -472,7 +468,6 @@ export GLOO_SOCKET_IFNAME=<network-interface>
export HCCL_OP_EXPANSION_MODE=AIV
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -566,7 +561,6 @@ export HCCL_OP_EXPANSION_MODE=AIV
export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=20
export SGLANG_SCHEDULER_DECREASE_PREFILL_IDLE=1
export SGLANG_SET_CPU_AFFINITY=1
@@ -660,7 +654,6 @@ export GLOO_SOCKET_IFNAME=<network-interface>
export HCCL_OP_EXPANSION_MODE=AIV
export HCCL_SOCKET_IFNAME=<network-interface>
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=100
export SGLANG_SCHEDULER_DECREASE_PREFILL_IDLE=1
export SGLANG_SET_CPU_AFFINITY=1
docs_new/docs/hardware-platforms/ascend-npus/best_practice/qwen3_6_35b_a3b.mdx
-9
View File
@@ -82,7 +82,6 @@ export HCCL_OP_EXPANSION_MODE=AIV
export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=30
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -180,7 +179,6 @@ export HCCL_OP_EXPANSION_MODE=AIV
export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=10
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -280,7 +278,6 @@ export HCCL_OP_EXPANSION_MODE=AIV
export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=20
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -377,7 +374,6 @@ export HCCL_OP_EXPANSION_MODE=AIV
export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=30
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -480,7 +476,6 @@ export HCCL_OP_EXPANSION_MODE=AIV
export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -576,7 +571,6 @@ export HCCL_OP_EXPANSION_MODE=AIV
export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=0
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -673,7 +667,6 @@ export HCCL_OP_EXPANSION_MODE=AIV
export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_PREFILL_DELAYER_MAX_DELAY_PASSES=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -773,7 +766,6 @@ export HCCL_OP_EXPANSION_MODE=AIV
export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=0
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
@@ -876,7 +868,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN=1
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_SET_CPU_AFFINITY=1
export STREAMS_PER_DEVICE=32
docs_new/docs/hardware-platforms/ascend-npus/best_practice/qwen3_next_80b_a3b_instruct.mdx
-3
View File
@@ -82,7 +82,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=400
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_ENABLE_TP_MEMORY_INBALANCE_CHECK=0
export SGLANG_NPU_USE_MULTI_STREAM=0
export SGLANG_SET_CPU_AFFINITY=1
@@ -194,7 +193,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=330
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_ENABLE_TP_MEMORY_INBALANCE_CHECK=0
export SGLANG_NPU_USE_MULTI_STREAM=0
export SGLANG_SET_CPU_AFFINITY=1
@@ -312,7 +310,6 @@ export HCCL_SOCKET_IFNAME=<network-interface>
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=400
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_ENABLE_TP_MEMORY_INBALANCE_CHECK=0
export SGLANG_NPU_USE_MULTI_STREAM=0
export SGLANG_WARMUP_TIMEOUT=3600
docs_new/docs/hardware-platforms/ascend-npus/model-tutorials/glm_5_2.mdx
-3
View File
@@ -200,7 +200,6 @@ source /usr/local/Ascend/nnal/atb/set_env.sh
export STREAMS_PER_DEVICE=32
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=600
export SGLANG_ENABLE_SPEC_V2=1
# MTP OVERLAP
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_NPU_USE_MULTI_STREAM=1
@@ -332,7 +331,6 @@ source /usr/local/Ascend/nnal/atb/set_env.sh
export STREAMS_PER_DEVICE=32
export SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT=600
# MTP OVERLAP
export SGLANG_ENABLE_SPEC_V2=1
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_NPU_USE_MULTI_STREAM=1
@@ -478,7 +476,6 @@ do
echo "${D_IP[$i]}"
export SGLANG_ENABLE_OVERLAP_PLAN_STREAM=1
export SGLANG_ENABLE_SPEC_V2=1
export HCCL_BUFFSIZE=650
export SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=32
docs_new/src/snippets/autoregressive/deepseek-r1-basic-deployment.jsx
-1
View File
@@ -122,7 +122,6 @@ export const DeepSeekR1BasicDeployment = () => {
command += ' \\\n --ep 8';
}
if (strategyValues.includes('mtp')) {
command = 'SGLANG_ENABLE_SPEC_V2=1 ' + command;
command +=
' \\\n --speculative-algorithm EAGLE' +
' \\\n --speculative-num-steps 3' +
docs_new/src/snippets/autoregressive/deepseek-v3-deployment.jsx
-1
View File
@@ -154,7 +154,6 @@ export const DeepSeekV3Deployment = () => {
if (strategyArray.includes('dp')) cmd += ' \\\n --dp 8 \\\n --enable-dp-attention';
if (strategyArray.includes('ep')) cmd += ' \\\n --ep 8';
if (strategyArray.includes('mtp')) {
cmd = 'SGLANG_ENABLE_SPEC_V2=1 ' + cmd;
cmd += ' \\\n --speculative-algorithm EAGLE \\\n --speculative-num-steps 3 \\\n --speculative-eagle-topk 1 \\\n --speculative-num-draft-tokens 4';
}
docs_new/src/snippets/autoregressive/glm-45-deployment.jsx
-1
View File
@@ -127,7 +127,6 @@ export const GLM45Deployment = () => {
cmd += ` \\\n --ep 8`;
}
if (strategyArray.includes('mtp')) {
cmd = 'SGLANG_ENABLE_SPEC_V2=1 ' + cmd;
cmd += ` \\\n --speculative-algorithm EAGLE \\\n --speculative-num-steps 3 \\\n --speculative-eagle-topk 1 \\\n --speculative-num-draft-tokens 4`;
}
docs_new/src/snippets/autoregressive/glm-46-deployment.jsx
-1
View File
@@ -141,7 +141,6 @@ export const GLM46Deployment = () => {
cmd += ` \\\n --ep 8`;
}
if (strategyArray.includes('mtp')) {
cmd = 'SGLANG_ENABLE_SPEC_V2=1 ' + cmd;
cmd += ` \\\n --speculative-algorithm EAGLE \\\n --speculative-num-steps 3 \\\n --speculative-eagle-topk 1 \\\n --speculative-num-draft-tokens 4`;
}
docs_new/src/snippets/autoregressive/glm-47-deployment.jsx
-1
View File
@@ -203,7 +203,6 @@ export const GLM47Deployment = () => {
// MTP / EAGLE speculative decoding (all platforms)
if (strategyArray.includes('mtp')) {
cmd = 'SGLANG_ENABLE_SPEC_V2=1 ' + cmd;
cmd += ` \\\n --speculative-algorithm EAGLE \\\n --speculative-num-steps 3 \\\n --speculative-eagle-topk 1 \\\n --speculative-num-draft-tokens 4`;
}
docs_new/src/snippets/autoregressive/glm-47-flash-deployment.jsx
-1
View File
@@ -117,7 +117,6 @@ export const GLM47FlashDeployment = () => {
cmd += ` \\\n --dp 1 \\\n --enable-dp-attention`;
}
if (strategyArray.includes('mtp')) {
cmd = 'SGLANG_ENABLE_SPEC_V2=1 ' + cmd;
if (hardware === 'b200') {
cmd += ` \\\n --speculative-draft-attention-backend triton`;
docs_new/src/snippets/autoregressive/gpt-oss-deployment.jsx
+1 -1
View File
@@ -152,7 +152,7 @@ export const GPTOSSDeployment = () => {
}
if (speculative === 'enabled') {
cmd += 'SGLANG_ENABLE_SPEC_V2=1 SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN=1 ';
cmd += 'SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN=1 ';
}
cmd += 'python -m sglang.launch_server \\\n';
docs_new/src/snippets/autoregressive/hunyuan3-preview-deployment.jsx
-1
View File
@@ -117,7 +117,6 @@ export const Hunyuan3PreviewDeployment = () => {
const enableSpec = values.speculative === 'enabled' && !isXeon;
let cmd = '';
if (enableSpec) cmd += 'SGLANG_ENABLE_SPEC_V2=1 ';
cmd += 'sglang serve \\\n';
cmd += ` --model-path ${modelName}`;
cmd += ` \\\n --tp ${tpValue}`;
docs_new/src/snippets/autoregressive/intern-s2-preview-deployment.jsx
+1 -2
View File
@@ -87,8 +87,7 @@ export const InternS2PreviewDeployment = () => {
flags.push(' --host 0.0.0.0');
flags.push(' --port 30000');
const envPrefix = mtp === 'enabled' ? 'SGLANG_ENABLE_SPEC_V2=1 \\\n' : '';
return `${envPrefix}sglang serve \\\n${flags.join(' \\\n')}`;
return `sglang serve \\\n${flags.join(' \\\n')}`;
};
const containerStyle = { maxWidth: '900px', margin: '0 auto', display: 'flex', flexDirection: 'column', gap: '4px' };
docs_new/src/snippets/autoregressive/mimo-v2-flash-deployment.jsx
+3 -3
View File
@@ -101,11 +101,11 @@ export const MiMoV2FlashDeployment = () => {
}
const commandPrefix = isMI355X
? 'PYTHONPATH=/sgl-workspace/aiter SGLANG_USE_AITER=0 USE_ROCM_AITER_ROPE_BACKEND=0'
: 'SGLANG_ENABLE_SPEC_V2=1';
? 'PYTHONPATH=/sgl-workspace/aiter SGLANG_USE_AITER=0 USE_ROCM_AITER_ROPE_BACKEND=0 '
: '';
const tpSize = isMI355X ? 4 : 8;
let cmd = `${commandPrefix} sglang serve \\\n`;
let cmd = `${commandPrefix}sglang serve \\\n`;
cmd += ` --model-path ${modelPath} \\\n`;
cmd += ` --trust-remote-code \\\n`;
cmd += ` --tp-size ${tpSize}`;
docs_new/src/snippets/autoregressive/mimo-v25-deployment.jsx
+1 -2
View File
@@ -15,7 +15,7 @@ export const MiMoV25Deployment = () => {
// GB300 → tp=4, dp=1, single-node, FP8 (Blackwell: vision fa4)
//
// Optional toggles:
// EAGLE MTP — adds --speculative-* flags + SGLANG_ENABLE_SPEC_V2=1.
// EAGLE MTP — adds --speculative-* flags.
// DeepEP — Hopper only (Blackwell uses flashinfer_trtllm). Adds
// --moe-a2a-backend deepep + --moe-dense-tp-size 1
// (and --ep on Pro) + SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=256.
@@ -300,7 +300,6 @@ export const MiMoV25Deployment = () => {
if (isPro && blackwell && multinode) {
envVars.push("NCCL_MNNVL_ENABLE=1", "NCCL_CUMEM_ENABLE=1");
}
if (useMtp) envVars.push("SGLANG_ENABLE_SPEC_V2=1");
if (useDeepep) envVars.push("SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=256");
// ---- flags ----
docs_new/src/snippets/autoregressive/nemotron3-super-deployment.jsx
+1 -2
View File
@@ -79,8 +79,7 @@ export const Nemotron3SuperDeployment = () => {
const modelPath = MODEL_PATHS[model] || MODEL_PATHS['bf16'];
const specV2Env = values.mtp === 'enabled' ? 'SGLANG_ENABLE_SPEC_V2=1 ' : '';
let cmd = `${specV2Env}sglang serve \\\n`;
let cmd = `sglang serve \\\n`;
cmd += ` --model-path ${modelPath} \\\n`;
cmd += ` --trust-remote-code \\\n`;
cmd += ` --tp ${tp} \\\n`;
docs_new/src/snippets/autoregressive/qwen36-deployment.jsx
+3 -8
View File
@@ -177,8 +177,8 @@ export const Qwen36Deployment = () => {
// NVFP4: nvidia/Qwen3.6-27B-NVFP4 on Blackwell (B200/B300). Follows the exact command
// shape from the checkpoint's docs — explicit --tp-size 1, --attention-backend trtllm_mha,
// new-style --mamba-radix-cache-strategy, and explicit --host/--port (no --mem-fraction-static
// or SGLANG_ENABLE_SPEC_V2 prefix). Reasoning / tool-call parsers still follow their toggles.
// new-style --mamba-radix-cache-strategy, and explicit --host/--port (no
// --mem-fraction-static). Reasoning / tool-call parsers still follow their toggles.
if (quantization === 'nvfp4') {
let cmd = `sglang serve --model-path nvidia/Qwen3.6-${sizeConfig.baseName}-NVFP4`;
cmd += ` \\\n --tp-size ${hwConfig.tp} --attention-backend trtllm_mha`;
@@ -199,12 +199,7 @@ export const Qwen36Deployment = () => {
const quantSuffix = quantization === 'fp8' ? '-FP8' : '';
const modelName = `Qwen/Qwen3.6-${sizeConfig.baseName}${quantSuffix}`;
let cmd = '';
if (speculative === 'enabled') {
cmd += 'SGLANG_ENABLE_SPEC_V2=1 ';
}
cmd += `sglang serve --model-path ${modelName}`;
let cmd = `sglang serve --model-path ${modelName}`;
if (hardware === 'xeon') {
cmd += ` \\\n --device cpu \\\n --disable-overlap-schedule`;
}