要在AWS Batch中始终为每个作业启动新的EC2实例，您可以使用EC2资源类型的Compute Environment，并将实例类型设置为EC2实例。以下是一个代码示例，说明如何创建一个具有此配置的AWS Batch作业定义。

import boto3

# 创建AWS Batch作业定义
def create_job_definition():
    batch_client = boto3.client('batch')
    
    # 定义作业容器的属性
    container_properties = {
        'image': 'your_docker_image',
        'vcpus': 2,
        'memory': 4096,
        'command': ['your_script.py']
    }
    
    # 创建作业定义
    response = batch_client.register_job_definition(
        jobDefinitionName='your_job_definition',
        type='container',
        containerProperties=container_properties,
        retryStrategy={
            'attempts': 1
        }
    )
    
    return response['jobDefinitionArn']

# 创建AWS Batch作业队列
def create_job_queue():
    batch_client = boto3.client('batch')
    
    # 创建计算环境
    response = batch_client.create_compute_environment(
        computeEnvironmentName='your_compute_environment',
        type='EC2',
        state='ENABLED',
        computeResources={
            'type': 'EC2',
            'minvCpus': 0,
            'maxvCpus': 10,
            'instanceTypes': ['your_instance_type'],
            'instanceRole': 'your_instance_role',
            'subnets': ['your_subnet_id'],
            'securityGroupIds': ['your_security_group_id']
        }
    )
    
    # 创建作业队列
    response = batch_client.create_job_queue(
        jobQueueName='your_job_queue',
        state='ENABLED',
        priority=1,
        computeEnvironmentOrder=[
            {
                'order': 1,
                'computeEnvironment': 'your_compute_environment'
            }
        ]
    )
    
    return response['jobQueueArn']

# 提交AWS Batch作业
def submit_job():
    batch_client = boto3.client('batch')
    
    response = batch_client.submit_job(
        jobName='your_job',
        jobQueue='your_job_queue',
        jobDefinition='your_job_definition',
        parameters={
            'param1': 'value1',
            'param2': 'value2'
        }
    )
    
    return response['jobId']

# 主函数
def main():
    job_definition_arn = create_job_definition()
    job_queue_arn = create_job_queue()
    job_id = submit_job()
    
    print('Job submitted. Job ID: {}'.format(job_id))

if __name__ == '__main__':
    main()

请注意，上述代码示例仅提供了一个基本的框架，并假设您已经设置了AWS凭证和适当的权限。