Restore optimization/scheduler states saved without pytorch-lightning

[ewrfcas]

If I have pre-trained a model (such as VIT) with its model and optimization state without saving by pytorch-lightning, how to restore the optimization for pytorch-lightning?
The model weights can be easily loaded by load_state_dict, but how to access optimization and further retore the state of it?

[ewrfcas]

Here is my solution, but I think it is not elegant.

model = LightningModule(...)
model.opt_state = None
model.opt_state = torch.load(ckpt_path)['opt']

Then for the LightningModule model:

def configure_optimizers(self):
      opt = Adam(...)
      if self.opt_state is not None:
            opt.load_state_dict(self.opt_state)
     return opt

I am looking for a better solution.