Pytorch-lightning: load_from_checkpoint raises TypeError when **kwargs is provided

[Guitaricet]

🐛 Bug

The issue happens in case of .save_hyperparameters() is not used.

To Reproduce

Steps to reproduce the behavior: follow code sample

Code sample

import os

import torch
from torch.nn import functional as F
from torch.utils.data import DataLoader
from torchvision.datasets import MNIST
from torchvision import transforms

import pytorch_lightning as pl
from pytorch_lightning.core.lightning import LightningModule

class LitModel(LightningModule):

    def __init__(self, some_stuff):
        super().__init__()
        self.l1 = torch.nn.Linear(28 * 28, 10)

    def forward(self, x):
        return torch.relu(self.l1(x.view(x.size(0), -1)))

    def training_step(self, batch, batch_idx):
        x, y = batch
        y_hat = self(x)
        loss = F.cross_entropy(y_hat, y)
        tensorboard_logs = {'train_loss': loss}
        return {'loss': loss, 'log': tensorboard_logs}

    def configure_optimizers(self):
        return torch.optim.Adam(self.parameters(), lr=0.001)

    def train_dataloader(self):
        dataset = MNIST(os.getcwd(), train=True, download=True, transform=transforms.ToTensor())
        loader = DataLoader(dataset, batch_size=32, num_workers=4, shuffle=True)
        return loader


model = LitModel(42)
trainer = pl.Trainer(fast_dev_run=True)
trainer.fit(model)

trainer.save_checkpoint('checkpoint.pl')
model = LitModel.load_from_checkpoint('checkpoint.pl', some_stuff=42)

# TypeError: __init__() got multiple values for argument 'some_stuff'

Expected behavior

Model is able to load

Environment

• PyTorch Version (e.g., 1.0): 1.5.0
• OS (e.g., Linux): Linux
• How you installed PyTorch (conda, pip, source): pip
• Build command you used (if compiling from source):
• Python version: 3.7.6
• CUDA/cuDNN version: N/A
• GPU models and configuration: N/A
• Any other relevant information:

Additional context

I noticed that loading starts working if you call save_hyperparameters. It also works correctly if some_stuff is not inside save_hyperparameters, e.g. if you change __init__ to this one

    def __init__(self, some_stuff, lr=42):
        super().__init__()
        self.l1 = torch.nn.Linear(28 * 28, 10)
        self.lr = 42
        self.save_hyperparameters('lr')

[github-actions[bot]]

Hi! thanks for your contribution!, great first issue!

[Borda]

ok, it seems a bit tricky how to correctly interpret method arguments and still keep back-compatibility...

[Guitaricet]

Maybe let's wait until the next major version to fix it, but document the behaviour.
It's not a bug, if it is documented, right? 😆

And in my (reasonably complex) case it was resolved when I added .save_hyperparameters() at the end of __init__. No other code changes were needed.

[rohitgr7]

Similar #2550?

[Guitaricet]

Yes, exactly the same issue

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[edenlightning]

@Borda please check if issue persists on master