pytorch-lightning: LR finder broken

#614 🐛 Bug

To Reproduce

Steps to reproduce the behavior:

model = TestModel()
trainer = pl.Trainer(gpus=1, default_save_path=exp_path, max_epochs=100)   

def configure_optimizers(self):
        optim = torch.optim.Adam(self.parameters(), lr=self.lr)
        sched = torch.optim.lr_scheduler.ReduceLROnPlateau(optim, 'min')
        return [optim], [sched]

# Run learning rate finder
lr_finder = trainer.lr_find(model)
# Results can be found in
lr_finder.results
# Plot with
fig = lr_finder.plot(suggest=True)
fig.show()

The following returns consistently:

 optimizer got an empty parameter list

The regular .fit method works as expected.

PL version: ‘0.7.6’