Added LoCON from LyCORIS
This commit is contained in:
@@ -93,6 +93,7 @@ class SDTrainer(BaseSDTrainProcess):
|
||||
|
||||
# back propagate loss to free ram
|
||||
loss.backward()
|
||||
torch.nn.utils.clip_grad_norm_(self.params, self.train_config.max_grad_norm)
|
||||
flush()
|
||||
|
||||
# apply gradients
|
||||
|
||||
Reference in New Issue
Block a user