Add distributed mode benchmark

.github/workflows/benchmark.yml

@@ -0,0 +1,33 @@
+name: AgentScope Benchmark
+
+on: [push, pull_request]
+permissions:
+  contents: write
+
+jobs:
+  bench:
+    permissions:
+      contents: write
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: True
+      matrix:
+        os: [ubuntu-latest]
+    env:
+      OS: ${{ matrix.os }}
+      PYTHON: '3.9'
+    steps:
+    - uses: actions/checkout@master
+    - name: Setup Python
+      uses: actions/setup-python@master
+      with:
+        python-version: 3.9
+    - name: Install AgentScope
+      run: |
+        pip install -q -e .[distribute]
+        pip install pytest pytest-benchmark
+    - name: Run bench
+      run: |
+        cd examples/distributed_simulation
+        pytest bench.py --benchmark-json output.json
+        cat output.json| jq -r ".benchmarks[] | .stats"

examples/distributed_simulation/bench.py

@@ -0,0 +1,49 @@
+# -*- coding: utf-8 -*-
+"""Simulation benchmark
+
+Please install pytest and pytest-benchmark to run this benchmark.
+"""
+
+import pytest  # pylint: disable=W0611
+from participant import Moderator, RandomParticipant, LLMParticipant
+from main import run_main_process  # pylint: disable=E0611
+import agentscope
+from agentscope.server import RpcAgentServerLauncher
+
+
+def test_simulation(benchmark):  # type: ignore[no-untyped-def]
+    """A single benchmark for the simulation"""
+    base_port = 23300
+    par_server_num = 4
+    mod_server_num = 1
+    agentscope.init(
+        project="simulation",
+        name="server",
+        save_code=False,
+        save_api_invoke=False,
+        model_configs="configs/model_configs.json",
+        use_monitor=False,
+    )
+    launchers = []
+    for i in range(par_server_num + mod_server_num):
+        launcher = RpcAgentServerLauncher(
+            host="localhost",
+            port=base_port + i,
+            custom_agents=[Moderator, RandomParticipant, LLMParticipant],
+        )
+        launcher.launch()
+        launchers.append(launcher)
+    benchmark(
+        run_main_process,
+        hosts=["localhost"],
+        base_port=23300,
+        server_per_host=4,
+        model_per_host=1,
+        participant_num=400,
+        moderator_per_host=1,
+        agent_type="random",
+        max_value=100,
+        sleep_time=1.0,
+    )
+    for launcher in launchers:
+        launcher.shutdown()