AttributeError in distributed mode -- (Avoid type conversion outside worker)

[rayrayraykk]

Describe the bug
When model contains BN layer, the param bn.num_batches_tracked would be convert to int by grpc. But the trainer.update can't handle this situation well.

A dummy solution:

  def update(self, model_parameters):
      '''
          Called by the FL client to update the model parameters
      Arguments:
          model_parameters (dict): PyTorch Module object's state_dict.
      '''
      for key in model_parameters:
          if isinstance(model_parameters[key], list):
              model_parameters[key] = torch.FloatTensor(
                  model_parameters[key])
          elif isinstance(model_parameters[key], int):
              model_parameters[key] = torch.tensor(model_parameters[key], dtype=torch.long)
              print(key, model_parameters[key])
          elif isinstance(model_parameters[key], float):
              model_parameters[key] = torch.tensor(model_parameters[key], dtype=torch.float)
      self.ctx.model.load_state_dict(self._param_filter(model_parameters),
                                     strict=False)

or can we solve it before sending the model_param?

[rayrayraykk]

Avoid type conversion outside worker

There is another type conversion bug in aggregator.

If the value is int, like 8, FloatTensor(8), would convert 8 to a random tensor with shape 8.
Should we handle these two situation in the message buffer instead of in aggregator?

Need a discussion @xieyxclack @joneswong .

A dummy solution in aggregator:

  if not self.cfg.federate.use_ss:
      if isinstance(local_model[key], torch.Tensor):
          local_model[key] = local_model[key].float()
      elif isinstance(local_model[key], list):
          local_model[key] = torch.FloatTensor(local_model[key])
      elif isinstance(local_model[key], int):
          local_model[key] = torch.tensor(local_model[key], dtype=torch.long)
      elif isinstance(local_model[key], float):
          local_model[key] = torch.tensor(local_model[key], dtype=torch.float)

[xieyxclack]

Avoid type conversion outside worker

IMO, the type conversion cannot be happened inside the worker, since the worker should be language-independent