r/stable-diffusion-finetune
1
0
Fork 0
Code Issues Pull requests Projects Releases Wiki Activity

reduce lr for ucg training

Browse source
This commit is contained in:
Patrick Esser 2022-07-29 10:41:44 +00:00
parent 9fd981d790
commit 1d75252371
1 changed files with 3 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

3
scripts/slurm/v1_iahr_torch111_ucg/launcher.sh
View file

@ -35,6 +35,9 @@ EXTRA="${EXTRA} data.params.min_size=512 data.params.max_pwatermark=0.4999"
# unconditional guidance training
EXTRA="${EXTRA} model.params.ucg_training.txt.p=0.1 model.params.ucg_training.txt.val=''"
# reduce lr a bit
EXTRA="${EXTRA} model.params.scheduler_config.params.f_max=[0.5]"
# postfix
EXTRA="${EXTRA} -f v1_iahr_torch111_ucg"