54 lines
1.1 KiB
Plaintext
54 lines
1.1 KiB
Plaintext
|
#!/bin/bash
|
||
|
|
|
||
|
|
#SBATCH --job-name=vel2vel
|
||
|
|
#SBATCH --dependency=singleton
|
||
|
|
#SBATCH --output=%x-%j.out
|
||
|
|
#SBATCH --error=%x-%j.err
|
||
|
|
|
||
|
|
#SBATCH --partition=gpu
|
||
|
|
#SBATCH --gres=gpu:v100-32gb:4
|
||
|
|
|
||
|
|
#SBATCH --exclusive
|
||
|
|
#SBATCH --nodes=2
|
||
|
|
#SBATCH --mem=0
|
||
|
|
#SBATCH --time=2-00:00:00
|
||
|
|
|
||
|
|
|
||
|
|
hostname; pwd; date
|
||
|
|
|
||
|
|
|
||
|
|
module load gcc openmpi2
|
||
|
|
module load cuda/10.1.243_418.87.00 cudnn/v7.6.2-cuda-10.1
|
||
|
|
|
||
|
|
source $HOME/anaconda3/bin/activate torch
|
||
|
|
|
||
|
|
|
||
|
|
export MASTER_ADDR=$HOSTNAME
|
||
|
|
export MASTER_PORT=8888
|
||
|
|
|
||
|
|
|
||
|
|
data_root_dir="/mnt/ceph/users/yinli/Quijote"
|
||
|
|
|
||
|
|
in_dir="linear"
|
||
|
|
tgt_dir="nonlin"
|
||
|
|
|
||
|
|
train_dirs="*[1-9]"
|
||
|
|
val_dirs="*[1-9]0"
|
||
|
|
|
||
|
|
files="vel/128x???.npy"
|
||
|
|
in_files="$files"
|
||
|
|
tgt_files="$files"
|
||
|
|
|
||
|
|
|
||
|
|
srun m2m.py train \
|
||
|
|
--train-in-patterns "$data_root_dir/$in_dir/$train_dirs/$in_files" \
|
||
|
|
--train-tgt-patterns "$data_root_dir/$tgt_dir/$train_dirs/$tgt_files" \
|
||
|
|
--val-in-patterns "$data_root_dir/$in_dir/$val_dirs/$in_files" \
|
||
|
|
--val-tgt-patterns "$data_root_dir/$tgt_dir/$val_dirs/$tgt_files" \
|
||
|
|
--in-channels 3 --out-channels 3 --norms cosmology.vel --augment \
|
||
|
|
--epochs 128 --batches-per-gpu 4 --loader-workers-per-gpu 4
|
||
|
|
# --load-state checkpoint.pth
|
||
|
|
|
||
|
|
|
||
|
|
date
|