Exception: You are trying to self.log() but it is not managed by the Trainer control flow

[kamyarghajar]

Hi,

I am using PyTorch Lightning and I do not know how can I fix this error. Can anyone help me with it because I could not find a proper answer for it:

pytorch_lightning.utilities.exceptions.MisconfigurationException: You are trying to `self.log()` but it is not managed by the `Trainer` control flow

And this is how I am instatiating the Trainer:

model = Classifier(conf)
tb_logger = TensorBoardLogger(
        save_dir=conf.outputs,
        version='version_' + datetime.now().strftime('%d-%m-%Y--%H-%M-%S'),
        name='pl_logs'
)
ckpt_path = os.path.join(conf.outputs, str(tb_logger.version), 'checkpoints')
checkpoint_callback = ModelCheckpoint(
     dirpath=ckpt_path,
     save_top_k=conf.save_top_k,
     verbose=True,
     monitor=conf.monitor,
     mode=conf.metric_mode,
     save_weights_only=True,
  )
early_stop_callback = EarlyStopping(
        monitor=conf.monitor,
        min_delta=0.0,
        patience=conf.patience,
        verbose=True,
        mode=conf.metric_mode,
 )
trainer = Trainer(
        logger=tb_logger,
        enable_checkpointing=True,
        callbacks=[
            checkpoint_callback,
            early_stop_callback
        ],
        gradient_clip_val=1.0,
        gpus=conf.gpus,
        log_every_n_steps=1,
        log_gpu_memory='all',
        deterministic=True,
        check_val_every_n_epoch=1,
        fast_dev_run=False,
        accumulate_grad_batches=conf.accumulate_grad_batches,
        max_epochs=conf.max_epochs,
        min_epochs=conf.min_epochs,
        val_check_interval=conf.val_check_interval,
        strategy='ddp',
)

[akihironitta]

@kamyarghajar As the error message states, you'll see the error because self.log(...) is called in places that the trainer has no control. Could you paste your full error message and script to see where exactly you're trying to call self.log(...)?

[kamyarghajar]

Thanks for the reply. Actually I just fixed it by adding some self.log calls in the training and validation step methods of my model module. I did not have any of such calls anywhere but I don’t know why the error was saying that I was. Was it the tensorboard logger itself?

[rohitgr7]

just to clarify, was it throwing this error when you had no self.log calls?

[kamyarghajar]

Yes exactly.

[rohitgr7]

can you reproduce it? using https://colab.research.google.com/github/Lightning-AI/lightning/blob/master/examples/pl_bug_report/bug_report_model.ipynb