meteoswiss / c4dl-lightningdl Goto Github PK

Dear Leinonen, Thank you for you kindly open source your code. I'm try to reproduce your figure in your paper, but currently occur one issue after runing the training.py. What does CPCH mean? It seem didn't appear in the your paper.

I want to express my gratitude for kindly open sourcing your code. I am currently studying your paper "Seamless lightning nowcasting with recurrent-convolutional deep learning" and I am trying to understand the details of your code. However, I have encountered a problem with the keys of the raw data, which are "dict_keys(['patches', 'patch_coords', 'patch_times', 'zero_patch_coords', 'zero_patch_times', 'zero_value'])". Could you please provide me with more information on what these keys represent?

Hi @jleinonen I’m wondering why my result is different with your results. training.py is used to get batch_gen, and then used the functions in the plots_lightning.py. Do you have any suggesions? Would you mind give a reproduced jupyter notebook for your results? That would be great help. Thank you in advance.

model-example-0.pdf
model-example-1.pdf
model-example-2.pdf

Thank you very much for sharing! When I try to call setup_batch_gen in training.py, it doesn't work. There is a break at dask.compute :
raw = {
var_name: dask.delayed(regions.load_patches)(fn)
for (var_name, fn) in files.items()
}
raw = dask.compute(raw, scheduler="processes")[0]
The final result shows:
concurrent.futures.process.BrokenProcessPool: A process in the process pool was terminated abruptly while the future was running or pending.
How can I fix it?