Ph  dddlmZddlmZmZdgZ d dej deedeeefdZy) )nn)ListOptionalpartition_modelNmodulebalancedevicescd}d}g}|D]k}g}t|D]}|j|||dz }||n||} |jtj|j | |dz }mtj|S)ab Partions the model accross multiple GPU devices. Given an :class:`nn.Sequential ` module, partitions the model across multiple GPU devices according the provided ``balance`` and ``devices``. Args: module (:class:`nn.Sequential `): Sequential model representing the pipe. balance (List[int]): List indicating the number of layers in each partition. devices (List[int], optional): List indicating the device to use for each partition. Defaults to ``range(len(balance))`` r)rangeappendr Sequentialto) rrr device_idxpipe_idx balanced_pipe num_layerslayersidevices pC:\Users\daisl\Desktop\realtime-object-detection\venv\Lib\site-packages\torch/distributed/pipeline/sync/utils.pyrrs(JHM z"A MM&* + MH# 'GJ4GR]]F366v>?a  ==- (()N) torchrtypingrr__all__rintrrrrsH!   (, )  )c )$s)$ )r