Method start_instance

CommonKG/llm_infer.py:30–35

启动ollama实例在特定GPU和端口上

(self, num, port)

        time.sleep(startup_delay)  # 等待所有实例启动

    def start_instance(self, num, port):
        """启动ollama实例在特定GPU和端口上"""
        # cmd = f"CUDA_VISIBLE_DEVICES={num} OLLAMA_HOST={self.base_url}:{port} ollama serve"
        cmd = f"OLLAMA_HOST={self.base_url}:{port} ollama serve"
        print("Running command:", cmd)
        # subprocess.Popen(cmd, shell=True, stderr=subprocess.DEVNULL, stdout=subprocess.DEVNULL)

    def get_available_instance(self):
        """使用轮询策略获取一个可用的实例"""