[QUESTION] Progress calss out of context manager

I’m trying to print the progress of a deep model training. Since I cannot directly modify the training loop I need to use a class CustomLogger method which is called at each iteration. This class stores the total length train_data_len of the loop and the called method on_batch_end. I would like to define a progress class attribute and update it everytime the method is called, like the following:

class CustomLogger(Callback):
    def __init__(
        self, train_data_len=None, eval_data_len=None, logging_step=50
    ):

        self.train_data_len = train_data_len
        self.eval_data_len = eval_data_len
        # define the progress class and add train and eval
        self.progress = Progress()
        self.train_p = self.progress.add_task("Train", total=self.train_data_len)
        self.eval_p = self.progress.add_task("Eval", total=self.eval_data_len)


    def on_epoch_end(self, loss: float, logs: Interaction, epoch: int):
        # at the end of the epoch reset the tasks progress
        self.train_p.reset(total=self.train_data_len)
        self.eval_p.reset(total=self.eval_data_len)

     # called on every loop iteration
    def on_batch_end(
        self, logs: Interaction, loss: float, batch_id: int, is_training: bool
    ): 
        # update the relative task 
            if is_training:
                self.progress.update(self.train_p, advance=1)
            else:
                self.progress.update(self.eval_p, advance=1)

[nicofirst1]

Indeed it was a Pycharm problem and the solution reported in a previous issue

[willmcgugan]

Ah. There is a note here about a potential fix. But you will probably get better results with a terminal.