Pavel Tarashkevich

You can comment out assertion macro, at your own risk.

You need to add `use_gpu` argument in `predict` function, then add the same [code block](https://github.com/Shivanandroy/simpleT5/blob/94ad898c428f48d13cb40d0acd237198475e80b3/simplet5/simplet5.py#L394-L400) as in `load_model` to set `self.device` attribute properly. @Shivanandroy

Tested out now and it seems to be working. Make sure to copy the whole token.

@marcklingen `LANGFUSE_LOG_LEVEL` and `LANGFUSE_LOG_FORMAT` can be used to control logs level and format, but it would be nice to have an option to completely turn off Langfuse logs.