KeyError: 'Trying to restore training state but checkpoint contains only the model. This is probably due to ModelCheckpoint.save_weights_only being set to True.'

[morestart]

this is my code:

checkpoint_callback = ModelCheckpoint(
    monitor='hmean',
    mode='max',
    dirpath='../weights',
    filename='DB-{epoch:02d}-{hmean:.2f}',
    save_last=True,
    save_weights_only=True,
)

trainer = pl.Trainer(
    # open this, must drop last
    benchmark=True,
    checkpoint_callback=True,
    gpus=[0],
    max_epochs=1200,
    min_epochs=300,
    logger=[logger],
    callbacks=[early_stop, checkpoint_callback],
    resume_from_checkpoint='../weights/DB-epoch=130-hmean=0.70.ckpt'
)

when i try resume train from checkpoint, i got this error: KeyError: 'Trying to restore training state but checkpoint contains only the model. This is probably due to ModelCheckpoint.save_weights_only being set to True.'

[rohitgr7]

if you set save_weights_only=True in ModelCheckpoint then it won't save optimizer/scheduler states in an ideal case. So assigning this checkpoint to resume training won't work because it needs to restore optimizer/scheduler state as well to actually resume it.