mindspore.parallel.rank_list_for_convert

View Source On Gitee
mindspore.parallel.rank_list_for_convert(rank_id, src_strategy_file=None, dst_strategy_file=None)[source]

List of original distributed checkpoint rank index for obtaining the target checkpoint of a rank_id during the distributed checkpoint conversion.

Parameters

  • rank_id (int) – The rank of which distributed checkpoint needs to be obtained after conversion.

  • src_strategy_file (str) – Name of source sharding strategy file which saved by mindspore.parallel.auto_parallel.AutoParallel(cell).save_param_strategy_file(file_path). when the src_strategy_file is None, it means that the source sharding strategy is without any sharing for each parameter. Default: None.

  • dst_strategy_file (str) – Name of destination sharding strategy file which saved by mindspore.parallel.auto_parallel.AutoParallel(cell).save_param_strategy_file(file_path). when the dst_strategy_file is None, it means that the destination sharding strategy is without any sharing for each parameter. Default: None.

Returns

List, the rank list required for converting the distributed checkpoint of rank_id.

Raises

  • ValueError – src_strategy_file or dst_strategy_file is incorrect.

  • TypeError – src_strategy_file or dst_strategy_file is not a string.

  • TypeError – rank_id is not an int.

Supported Platforms:

Ascend

Examples

>>> from mindspore.parallel import rank_list_for_convert
>>> rank_id = 0
>>> rank_list = rank_list_for_convert(rank_id, "./src_strategy.ckpt", "./dst_strategy.ckpt")
>>> checkpoint_files_map = {}
>>> for rank in rank_list:
...     checkpoint_files_map[rank] = "./pangu{}-100_2.ckpt".format(rank)