Poor opportunities for developing


I was very disappointed at the moment when mlflow forced me to use torch lightning.

Why, when trying to write a selected device to self.device, the error can’t set attribute pops up? Why is this field blackened?

In addition, there are many complaints about the documentation on the site. It is not complete and does not allow you to customize an extremely inconvenient trainer.

Pure PyTorch is best!!!

Hi ’ pl_is_bad’,

self.device is already set and can’t be changed. (As obvious) It refers to the device being used.

Not sure if judging a library by a single parameter is the best decision, but its up to you!