This is the last part of our three-part series checking out the cool stuff the DeepSeek team has done. We're going to wrap things up by focusing on how they train their models.