The invention discloses a resource scheduling method and device for a deep learning frame. The method comprises the steps of inquiring states of all deep learning work objects from a Kubernetes platform at predetermined time intervals; in response to the fact that the deep learning work objects of which the states conform to submitted resource request states are inquired from all the deep learningwork objects, submitting a resource request to the Kubernetes platform to schedule a physical machine where the Kubernetes platform is located and start a deep learning training task. By means of theresource scheduling method for the deep learning frame, resource distribution and release of the deep learning training task can be automatically processed completely.