1.0.2 doesn't log hyper parameters

[xiadingZ]

In pt 1.0.2, tensorboard can't log hyperparameters, it only show hp_metrics. and hparams.yaml is empty

this is my code aboud hparams:

class Video_Base(pl.LightningModule):
    def __init__(self, hparams, *args, **kwargs):
        super().__init__()
        self.hparams = hparams
        ...

[ndrplz]

I think you should replace

self.hparams = hparams

with

self.save_hyperparameters()

(see here)

related: #4179

[Borda]

well we are saving the initial arguments, so you should be calling self.save_hyperparameters()
this is not just the case of TB, it is for all loggers since #4163

the reccomended flow is:

class MyModule(pl.LightningModule):
    def __init__(self, arg1, arg2):
        super().__init__()
        self.save_hyperparameters()
        ...

[Borda]

eventually, we may permit this action if self.hparams = hparams is inside __init__ or just for the fist time is called...