forked from mrq/DL-Art-School
20 lines
704 B
Python
20 lines
704 B
Python
|
import torch
|
||
|
|
|
||
|
|
# Quick script that can be used to convert from pretrained SRFlow weights to the variants used in this repo. The only
|
||
|
|
# differences between the two is the variable naming conventions used by the RRDBNet. (FWIW this repo is using the
|
||
|
|
# more up-to-date names that conform to Python standards).
|
||
|
|
|
||
|
|
official_weight_file = 'SRFlow_CelebA_8X.pth'
|
||
|
|
output = 'CelebA_converted.pth'
|
||
|
|
|
||
|
|
sd = torch.load(official_weight_file)
|
||
|
|
sdp = {}
|
||
|
|
for k,v in sd.items():
|
||
|
|
k = k.replace('RRDB.RRDB_trunk', 'RRDB.body')
|
||
|
|
k = k.replace('.RDB', '.rdb')
|
||
|
|
k = k.replace('trunk_conv.', 'conv_body.')
|
||
|
|
k = k.replace('.upconv', '.conv_up')
|
||
|
|
k = k.replace('.HRconv', '.conv_hr')
|
||
|
|
sdp[k] = v
|
||
|
|
torch.save(sdp, output)
|