Save wandb config file as hyperparameter

[rienboonstoppel]

I was wondering if there is some way to save a WandB config file using the save_hyperparameters() call in the Trainer. I am using WandB for logging purposes and to perform parameter sweeps etc. I initialize a WandB run with a config file, which is passed on to the Trainer. And I would like to save that config to the checkpoints I save using ModelCheckpoint callback. I know WandB saves the config that is used for a certain run with it in the same folder, but I would like them to be in the checkpoint as well. Is there some way to do this? Because it looks like the WandB config isn't allowed in the save_hyperparameter() at the moment...
So shortened in pseudo code I have something like this:

A main file with:

def main(default_config): # default_config is read from a file
    ...
    wandb.init(config=default_config)
    ...
    config = wandb.config
    ...
    model = LitTrainer(netG=generator, netD=discriminator, config=config)
    trainer = pl.Trainer(...)
    trainer.fit(model)

And then a trainer file with:

class LitTrainer(pl.LightningModule):
    def __init__(self,
                 netG, netD,
                 config,
                 ):
    super().__init__()
    self.save_hyperparameters(config)
    ...

And in the rest of the trainer all kinds of parameters are an attribute of that config. But when I try to save that config using the save_hyperparameters, it returns the following: ValueError: Unsupported config type of <class 'wandb.sdk.wandb_config.Config'>.
And first trying to convert the config back to a dict before passing it to save_hyperparameters doesn't seem to work as well.
Does anyone know how to do this?

[rienboonstoppel]

Apparently it has something to do with the config from WandB. The issue is that WandB config is a managed object so it really can't cleanly be deepcopy'ed. Using the static_config from WandB is serializable and works just fine!