Isaac Godfried

You could create the model yourself e.g. `from flood_forecast.tranformer_xl.transformer_basic import SimpleTransformer` and then write your own training loop. Otherwise no.

Seems to be the batch_size in training_params that is necessary right now. Should update the docs about it.

We've tested on Python 3.10 and newer (that is what our CI is on). It might be related to Python 3.8 issues would be my guess

r

Could you give a sample of the data? You could train a model for each, concat them all together/do multi targets or you could use a series id that you...

Hello sorry for the delay. We do currently have [Docker containers](https://github.com/AIStream-Peelout/flow_docker) which you can use with Wandb to perform a distributed hyper-parameter sweep. IMO multi-GPU for a single model isn't...

Interesting, I've never really run into that problem before. Let me look into it. FF is built on top of PyTorch of course so it is hopefully it is something...

https://flow-forecast.atlassian.net/wiki/spaces/FF/pages/1178730497/Training+locally+using+the+command+line beginning here. This file outlines how to run a sweep without using a notebook also how to run a basic training script without using the notebook.

Yes I have the same problem. Now December and no updates.

@antonpolishko