o Tio@sDddlmZmZddlmZmZddlZdZdZGdddeZ dS))ABCabstractmethod)IterableTupleNmegatron huggingfacec@s0eZdZdZedeeeej ffddZ dS)CheckpointEngineBasea Abstract interface for checkpoint engines to implement. There is no ``__init__`` method here by design, since the creation of the checkpoint engine will happen outside the policy/engine code. The tradeoff being made here is that we will write different frontends for different checkpoint engines, but these frontends can be tailored to the specific checkpoint engine/model source needs. returncCsdS)a This method should create a generator of tuples of the form (name, parameter) for all parameters in the model. The name should be the fully qualified name of the parameter, and the parameter should be a torch.Tensor. The expected use of a checkpoint engine is the following: ```python for name, parameter in checkpoint_engine.parameters(): container_map.map_param(name, parameter) ``` For a concrete use example, see ``InferenceV2Policy``. N)selfr r a/home/ubuntu/.local/lib/python3.10/site-packages/deepspeed/inference/v2/checkpoint/base_engine.py parametersszCheckpointEngineBase.parametersN) __name__ __module__ __qualname____doc__rrrstrtorchTensorr r r r r rs "r) abcrrtypingrrrMEGATRON HUGGINGFACErr r r r s