Load weight from checkpoint with class that takes mandatory arguments

[Loris697]

Hi to all,

I'm quite new to GitHub, maybe this question is in the wrong place. If something is missing I will update it later.

I'm using a PL model: a class that takes two mandatory arguments and other optional:

class PLModel(pl.LightningModule): def __init__(self, name, model, ...)

I saved the weight using a callback during the training:

checkpoint_callback = ModelCheckpoint( monitor='balValAcc', dirpath=folder, filename= '{epoch:02d}-{balValAcc:.2f}', save_top_k=3, mode='max', )

But when I try to load the checkpoints in this way:

try: testModel = PLModel(model.name, model.model).load_from_checkpoint(checkpoint_path=mypath + '/' + weightName) except Exception as e: print(e)

I get:

init() missing 2 required positional arguments: 'name' and 'model'

The I tried to change the code to

try: testModel = PLModel(model.name, model.model).load_from_checkpoint(mypath + '/' + weightName, model.name, model.model) except Exception as e: print(e)

but i get another error message:

don't know how to restore data location of torch.FloatStorage (tagged with EfficientNetB7)

Am I missing something?