add option for smooth l1 (huber / delta)

Author: kabachuha

library/train_util.py

@@ -3091,7 +3091,7 @@ def add_training_arguments(parser: argparse.ArgumentParser, support_dreambooth:
         "--loss_type",
         type=str,
         default="l2",
-        choices=["l2", "huber", "huber_scheduled"],
+        choices=["l2", "huber", "huber_scheduled", "smooth_l1", "smooth_l1_scheduled"],
         help="The type of loss to use and whether it's scheduled based on the timestep"
     )
     parser.add_argument(
@@ -4608,7 +4608,7 @@ def get_timesteps_and_huber_c(args, min_timestep, max_timestep, num_train_timest
 
     #TODO: if a huber loss is selected, it will use constant timesteps for each batch
     # as. In the future there may be a smarter way
-    if args.loss_type == 'huber_scheduled':
+    if args.loss_type == 'huber_scheduled' or args.loss_type == 'smooth_l1_scheduled': #NOTE: Will unify scheduled and vanilla soon
         timesteps = torch.randint(
             min_timestep, max_timestep, (1,), device='cpu'
         )
@@ -4617,7 +4617,7 @@ def get_timesteps_and_huber_c(args, min_timestep, max_timestep, num_train_timest
         alpha = - math.log(args.huber_c) / num_train_timesteps
         huber_c = math.exp(-alpha * timestep)
         timesteps = timesteps.repeat(b_size).to(device)
-    elif args.loss_type == 'huber':
+    elif args.loss_type == 'huber' or args.loss_type == 'smooth_l1':
         # for fairness in comparison 
         timesteps = torch.randint(
             min_timestep, max_timestep, (1,), device='cpu'
@@ -4670,6 +4670,12 @@ def conditional_loss(model_pred:torch.Tensor, target:torch.Tensor, reduction:str
             loss = torch.mean(loss)
         elif reduction == "sum":
             loss = torch.sum(loss)
+    elif loss_type == 'smooth_l1' or loss_type == 'smooth_l1_scheduled': # NOTE: Will unify in the next commits
+        loss = 2 * (torch.sqrt((model_pred - target) ** 2 + huber_c**2) - huber_c)
+        if reduction == "mean":
+            loss = torch.mean(loss)
+        elif reduction == "sum":
+            loss = torch.sum(loss)
     else:
         raise NotImplementedError(f'Unsupported Loss Type {loss_type}')
     return loss