Jacob Prince-Bieker

Just skim-read through this paper, it seems really relevant and has some of the same insights as in the graph paper! As in, the finer grained data, and more variables...

Not currently, I've been slowly pushing ERA5 data to Huggingface here https://huggingface.co/datasets/openclimatefix/era5 and the code for training with GFS should be fairly simple to change to use ERA5. Would welcome...

One great place would be HuggingFace Datasets, its free hosting, and has some nice features, as well as then you should be able to just take the GFS training script,...

Also NormalizedMSELoss currently doesn't take out the mean of the change like in the paper, it just ignores the variances as that was giving NaNs with the current implementation.

Thanks for the suggestion! Yeah, I can try adding that. I'm hoping to train these models on even more levels and variables than in the paper, and at a finer...

Hi, yeah that would be awesome!

Yeah, this is still of interest! It would be good to be able to use this on TPUs. Ideally still in PyTorch, although up for JAX as well.

Hi, just posting on here would work well!

The HuggingFace datasets script will be hosted here https://huggingface.co/datasets/openclimatefix/arco-era5

Yes, I believe so, I can assign you it?