diffae / run_ffhq256.py
qninhdt's picture
Upload 68 files
1ab03a3 verified
raw
history blame contribute delete
333 Bytes
from templates import *
from templates_latent import *
if __name__ == '__main__':
# 256 requires 8x v100s, in our case, on two nodes.
# do not run this directly, use `sbatch run_ffhq256.sh` to spawn the srun properly.
gpus = [0, 1, 2, 3]
nodes = 2
conf = ffhq256_autoenc()
train(conf, gpus=gpus, nodes=nodes)