Simplify initialization of Trainer

[clessig]

Describe the task. It can be a feature, documentation, etc.

run_train contain redundant code on init_torch(), init_ddp() etc:

WeatherGenerator/src/weathergen/run_train.py

Line 64 in 1776b0a

devices = Trainer.init_torch()

This is original since these functions should be called before the logger is initalized. The redundancy is a source of errors. init_torch(), init_ddp() also should be called in Trainer and not in run_train.py

Hedgedoc URL, if you are keeping notes, plots, logs in hedgedoc.

No response

Area

• datasets, data readers, data preparation and transfer
• model
• science
• infrastructure and engineering
• evaluation, export and visualization
• documentation

[tjhunter]

I would scope exactly the issue. The initialization of the model is very complex and this task, as described, could become a rabbit hole.