§ z”jú ãóp—ddlmZddlmZddlmZddlmZerddlmZddl m Z ddl m Z  ddd„Z dS)é)Ú annotations)Ú TYPE_CHECKING)Ú framework)Ústream)ÚTensor)Útask)ÚGroupNTÚtensorrÚ gather_listúlist[Tensor] | NoneÚdstÚintÚgroupú Group | NoneÚsync_opÚboolÚreturnú task | Nonecól—tj¦«s Jd¦«‚tj|||||¦«S)a Gather tensors from all participators. Args: tensor (Tensor): The input Tensor. Its data type should be float16, float32, float64, int32, int64, int8, uint8, bool or bfloat16. gather_list (list): A list of Tensors to hold the gathered tensors. Every element in the list must be a Tensor whose data type should be float16, float32, float64, int32, int64, int8, uint8, bool or bfloat16. Default value is None. dst (int): The dst rank id. Default value is 0. group (Group, optional): The group instance return by new_group or None for global default group. sync_op (bool, optional): Whether this op is a sync op. The default value is True. Returns: Async work handle,which can be wait on, if async_op is set to True. None, if not async_op Examples: .. code-block:: python >>> # doctest: +REQUIRES(env: DISTRIBUTED) >>> import paddle >>> import paddle.distributed as dist >>> dist.init_parallel_env() >>> gather_list = [] # type: ignore >>> if dist.get_rank() == 0: ... data = paddle.to_tensor([1, 2, 3]) ... dist.gather(data, gather_list, dst=0) >>> else: ... data = paddle.to_tensor([4, 5, 6]) ... dist.gather(data, gather_list, dst=0) >>> print(gather_list) >>> # [[1, 2, 3], [4, 5, 6]] (2 GPUs, out for rank 0) >>> # [] (2 GPUs, out for rank 1) z-gather doesn't support static graph mode yet.)rÚin_dynamic_moderÚgather)r r r rrs úw/lsinfo/ai/hellotax_ai/data_center/backend/venv/lib/python3.11/site-packages/paddle/distributed/communication/gather.pyrrsE€õV Ô $Ñ &Ô &ððØ7ñôÐ &õ Œ=˜ ¨c°5¸'Ñ BÔ BÐBó)NrNT) r rr r r rrrrrrr)Ú __future__rÚtypingrÚpaddlerÚ paddle.distributed.communicationrrÚpaddle.base.corerÚ&paddle.distributed.communication.groupr r©rrúr!sÅðð #Ð"Ð"Ð"Ð"Ð"à Ð Ð Ð Ð Ð àÐÐÐÐÐØ3Ð3Ð3Ð3Ð3Ð3àð=ØÐÐÐÐÐØ%Ð%Ð%Ð%Ð%Ð%Ø<Ð<Ð<Ð<Ð<Ð<ð (,ØØØð .Cð.Cð.Cð.Cð.Cð.Cð.Cr