Merge branch 'develop' into develop

Author: qw86972190

.github/workflows/_base_test.yml

@@ -206,6 +206,13 @@ jobs:
           check_service 90
           python -m pytest -sv test_max_waiting_time.py || TEST_EXIT_CODE=1
 
+          export TEMPLATE=TOKEN_NORMAL
+          curl -X POST http://0.0.0.0:${FLASK_PORT}/switch \
+            -H "Content-Type: application/json" \
+            -d "{\"--model\": \"/MODELDATA/ERNIE-4.5-VL-28B-A3B-Thinking\", \"--reasoning-parser\": \"ernie-45-vl-thinking\", \"--tool-call-parser\": \"ernie-45-vl-thinking\", \"--tensor-parallel-size\": 1, \"--quantization\": \"wint4\", \"--max-model-len\": 131072, \"--max-num-seqs\": 32}"
+          check_service 90
+          python -m pytest -sv test_prompt_ids.py || TEST_EXIT_CODE=1
+
           popd
           echo "TEST_EXIT_CODE=${TEST_EXIT_CODE}" >> /workspace/FastDeploy/exit_code.env
           '

benchmarks/README.md

@@ -58,7 +58,7 @@ python benchmark_serving.py \
   --port 9812 \
   --dataset-name EBChat \
   --dataset-path ./filtered_sharedgpt_2000_input_1136_output_200_fd.json \
-  --hyperparameter-path yaml/request_yaml/eb45t-32k.yaml \
+  --hyperparameter-path yaml/request_yaml/eb45-32k.yaml \
   --percentile-metrics ttft,tpot,itl,e2el,s_ttft,s_itl,s_e2el,s_decode,input_len,s_input_len,output_len \
   --metric-percentiles 80,95,99,99.9,99.95,99.99 \
   --num-prompts 1 \
@@ -78,7 +78,7 @@ python benchmark_serving.py \
   --port 9812 \
   --dataset-name EBChat \
   --dataset-path ./filtered_sharedgpt_2000_input_1136_output_200_fd.json \
-  --hyperparameter-path yaml/request_yaml/eb45t-32k.yaml \
+  --hyperparameter-path yaml/request_yaml/eb45-32k.yaml \
   --percentile-metrics ttft,tpot,itl,e2el,s_ttft,s_itl,s_e2el,s_decode,input_len,s_input_len,output_len \
   --metric-percentiles 80,95,99,99.9,99.95,99.99 \
   --num-prompts 2000 \
@@ -100,7 +100,7 @@ python benchmark_serving.py \
   --port 9812 \
   --dataset-name EBChat \
   --dataset-path ./filtered_sharedgpt_2000_input_1136_output_200_fd.json \
-  --hyperparameter-path yaml/request_yaml/eb45t-32k.yaml \
+  --hyperparameter-path yaml/request_yaml/eb45-32k.yaml \
   --percentile-metrics ttft,tpot,itl,e2el,s_ttft,s_itl,s_e2el,s_decode,input_len,s_input_len,output_len \
   --metric-percentiles 80,95,99,99.9,99.95,99.99 \
   --num-prompts 2000 \
@@ -135,3 +135,30 @@ python benchmarks/benchmark_mtp.py \
 --dataset-name：指定数据集类，指定为"EBChat"可读取转存的FD格式数据集
 --dataset-path：测试数据集路径
 ```
+
+### 指定输入输出长度，构造随机纯文输入测试
+
+相关参数：
+- --dataset-name：指定数据集类，指定为"random"可构造随机纯文输入
+- --random-input-len：随机输入长度，对应英文单词数，默认200
+- --random-output-len：随机输出长度，默认1024
+- --random-range-ratio：输入输出长度变化范围比，[length *(1 - range_ratio), length* (1 + range_ratio)]，默认0.1
+
+#### 使用方式：
+```bash
+python benchmark_serving.py \
+  --backend openai-chat \
+  --model EB45T \
+  --endpoint /v1/chat/completions \
+  --host 0.0.0.0 \
+  --port 9812 \
+  --dataset-name random \
+  --random-input-len 200 \
+  --random-output-len 1024 \
+  --random-range-ratio 0.1 \
+  --percentile-metrics ttft,tpot,itl,e2el,s_ttft,s_itl,s_e2el,s_decode,input_len,s_input_len,output_len \
+  --metric-percentiles 80,95,99,99.9,99.95,99.99 \
+  --num-prompts 2000 \
+  --max-concurrency 100 \
+  --save-result > infer_log.txt 2>&1 &
+```

benchmarks/backend_request_func.py

@@ -52,6 +52,7 @@ class RequestFuncInput:
     language: Optional[str] = None
     debug: bool = False
     response_format: Optional[dict] = None
+    random_flag: bool = False
 
 
 @dataclass
@@ -103,6 +104,13 @@ async def async_request_eb_openai_chat_completions(
         # 超参由yaml传入
         payload.update(request_func_input.hyper_parameters)
 
+        # 随机输入开关
+        if request_func_input.random_flag:
+            payload["max_tokens"] = request_func_input.output_len
+            metadata = payload.get("metadata", {})
+            metadata["min_tokens"] = request_func_input.output_len
+            payload["metadata"] = metadata
+
         if request_func_input.ignore_eos:
             payload["ignore_eos"] = request_func_input.ignore_eos