michael/stable-diffusion-webui
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

set TI AdamW default weight decay to 0

Browse source
This commit is contained in:
flamelaw 2022-11-27 00:35:44 +09:00
parent 1bd57cc979
commit 755df94b2a
1 changed files with 1 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
modules/textual_inversion/textual_inversion.py
View file

@ -283,7 +283,7 @@ def train_embedding(embedding_name, learn_rate, batch_size, gradient_step, data_
shared.sd_model.first_stage_model.to(devices.cpu)
embedding.vec.requires_grad = True
optimizer = torch.optim.AdamW([embedding.vec], lr=scheduler.learn_rate)
optimizer = torch.optim.AdamW([embedding.vec], lr=scheduler.learn_rate, weight_decay=0.0)
scaler = torch.cuda.amp.GradScaler()
batch_size = ds.batch_size