context, please.
what exactly do you want to replace ?
Looping in general is what I’m trying to avoid. I’m just looking for a silver bullet LOL
However, I did get numba working, using Python 3.9.11, numpy 1.21.5, and numba 0.55.1. It runs magnitudes of order faster now! See mc_gpu.py in the repository.
(runs at 60 fps realtime on my cheesy, 10 yrs old thinkpad)
Very nice find berak
I’ve calculated the meshes generated by AI and by ground truth. They are quite similar.
I have started Appendix C in my paper, to display Figure 15.
I tried using the CRUTEM5 weather station data. I am trying to train the PyTorch network to do regression, so that I can forecast the global warming from 2020 to 2050. The network doesn’t work. I am wondering if I’m training it properly. So, given two input dates (a range, or span of time), the network spits out the mean and standard deviation, in terms of degrees C per decade. The code is at GitHub - sjhalayka/crutem5_python
Sorry to necro a dead thread.