fix tp2

Author: gzy19990617

test/ci_use/QwQ-32B/run_server.sh

@@ -19,13 +19,13 @@ rm -rf log/*
 mkdir -p log
 
 # 环境变量
-export CUDA_VISIBLE_DEVICES=4,5,6,7
+export CUDA_VISIBLE_DEVICES=0,1
 export INFERENCE_MSG_QUEUE_ID=7679
 export ENABLE_V1_KVCACHE_SCHEDULER=1
 
 
 python -m fastdeploy.entrypoints.openai.api_server \
-       --tensor-parallel-size 4 \
+       --tensor-parallel-size 2 \
        --port 8787 \
        --engine-worker-queue-port 7679 \
        --metrics-port 7877 \

test/ci_use/QwQ-32B/run_test.sh

@@ -8,7 +8,7 @@ BASE_URL="http://$HOST:$PORT"
 
 TOTAL_ROUNDS=100
 CHAT_REQUESTS_PER_ROUND=5
-export CUDA_VISIBLE_DEVICES=4,5,6,7
+export CUDA_VISIBLE_DEVICES=0,1
 MAX_MEMORY_MB=10240  # 10GB
 
 # ====================================================