[Bug]: In PD decoupled deployment, using toy_proxy_server to send proxy requests will trigger an error when the `min_tokens` parameter is added, and the proxy service cannot process requests normally.

# jemalloc

export LD_PRELOAD=/usr/lib64/libjemalloc.so.2:$LD_PRELOAD
export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH

# Set the operator dispatch pipeline level to 1 and disable manual memory control in ACLGraph

export OMP_PROC_BIND=false
export OMP_NUM_THREADS=1
export TASK_QUEUE_ENABLE=1
export VLLM_ALLOW_BAD_CHECKPOINT=1

# Enable the AIVector core to directly schedule ROCE communication

export ASCEND_RT_VISIBLE_DEVICES=8,9,10,11,12,13,14,15

Reduce memory fragmentation and avoid out-of-memory errors.

export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True

export HCCL_OP_EXPANSION_MODE="AIV"
export HCCL_BUFFSIZE=1024
export OMP_NUM_THREADS=1
export TASK_QUEUE_ENABLE=1
echo performance | tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor
sysctl -w vm.swappiness=0
sysctl -w kernel.numa_balancing=0
sysctl kernel.sched_migration_cost_ns=50000

export HCCL_INTRA_ROCE_ENABLE=1
export HCCL_INTRA_PCIE_ENABLE=0
export VLLM_RPC_TIMEOUT=3600000
export VLLM_EXECUTE_MODEL_TIMEOUT_SECONDS=30000
export HCCL_EXEC_TIMEOUT=204
export HCCL_CONNECT_TIMEOUT=1200
export VLLM_ENGINE_READY_TIMEOUT_S=120000

nohup vllm serve /mnt/nfs_hw/weight/Qwen3.6-35B-A3B
--host 0.0.0.0
--port 13900
--data-parallel-size 1
--tensor-parallel-size 2
--seed 1024
--served-model-name qwen3.6
--max-num-seqs 128
--max-model-len 262144
--max-num-batched-tokens 16384
--trust-remote-code
--block-size 128
--dtype bfloat16
--gpu-memory-utilization 0.95
--enable-prefix-caching
--enable-chunked-prefill
--speculative-config '{"method": "qwen3_5_mtp", "num_speculative_tokens": 3, "enforce_eager": true}'
--compilation-config '{"cudagraph_mode":"FULL_DECODE_ONLY"}'
--additional-config '{"enable_cpu_binding":true}'
--async-scheduling
--reasoning-parser qwen3
--enable-auto-tool-choice
--tool-call-parser qwen3_coder
--kv-transfer-config
'{
"kv_connector": "UCMConnector",
"kv_role": "kv_both",
"kv_connector_module_path": "ucm.integration.vllm.ucm_connector",
"kv_connector_extra_config": {
"UCM_CONFIG_FILE": "/data/xpyd/ucm_config_hetero_pd.yaml"
}
}' \

/data/xpyd/qwen.log 2>&1 &

export CUDA_VISIBLE_DEVICES=6,7

nohup vllm serve /mnt/nfs_hw/weight/Qwen3.6-35B-A3B
--max-model-len 262144
--tensor-parallel-size 2
--served-model-name qwen3.6
--gpu_memory_utilization 0.87
--trust-remote-code
--port 13700
--block-size 128
--dtype bfloat16
--kv-transfer-config
'{
"kv_connector": "UCMConnector",
"kv_role": "kv_both",
"kv_connector_module_path": "ucm.integration.vllm.ucm_connector",
"kv_connector_extra_config": {
"UCM_CONFIG_FILE": "/data/xpyd/ucm_config_hetero_pd.yaml"
}
}' \

/data/xpyd/qwen.log 2>&1 &