File size: 771 Bytes
1739237 7798813 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 | # Example script to show the steps needed to downscale ACE
# using HiRO
# Provided are two inference configurations used in the
# manuscript:
# 1.) Global inference with ACE for 2023 then downscaled
# for most of the global (-66S to 70N) using HiRO
#
# 2.) Global inference with ACE for 2014 - 2023 then
# the Pacific NorthWest (PNW) only downscaled with HiRO
# First run coarse ACE2S
python -m fme.ace.inference ace2s_inference_config_global.yaml
# Second downscale ACE2S using HiRO
# HiRO is more computationally costly then ACE2S
# for faster through put more GPUs may be required
NGPU=1
# This will downscale a subselected region around the Pacific Northwest.
torchrun --nproc_per_node $NGPU -m fme.downscaling.inference hiro_downscaling_ace2s_pnw_output.yaml
|