Lukas

@zhiyuanpeng Thank you for pointing this out! Here's a proposition how to fix the issue: https://github.com/ashleve/lightning-hydra-template/pull/426

Hi @yangze68 , I don't know your use case but are you sure you need to use `torch.load()`? Lightning is built with other ways of loading checkpoint in mind, for...

I don't know if I understand correctly, but if you have two different folders/projects added to PYTHONPATH, both of which have the same folder structure (e.g. `src/models/...`), then the interpreter...

@turian `ddp_spawn` is not better but it's the only ddp mode that works correctly with hydra right now. As I mentioned, normal DDP generates multiple unwanted files. This is due...

@turian There will be more output directories as explained in https://github.com/facebookresearch/hydra/issues/2070  Just to make this clear, normal ddp actually computes correctly in hydra single run mode, but you will...

As the exception message says, lightning disables loggers in `fast_dev_run` mode so you shouldn't use it to debug or test wandb callbacks. There are however other flags you could use,...

@turian Checkpoints are always available in output dirs. Enabling uploading them as artifacts in wandb logger doesn't change that. If you really need it despite of that, perhaps you could...

@00krishna Thank you for using the repo! If you don't want to create a datamodule then simply don't do it - it should be easy to remove it from the...

> If you like I can share the code with you on adapting your template to the RL algorithms I found. Sure, if you want you can post it here...

This is a nice example (Lightning + Hydra + RL): https://github.com/belerico/lightning-rl