[bug fix] advanced dreambooth lora sdxl - fixes bugs described in #6486 (#6599)

* fixes bugs:
1. redundant retraction
2. param clone
3. stopping optimization of text encoder params

* param upscaling

* style

Author: linoytsaban

examples/advanced_diffusion_training/train_dreambooth_lora_sdxl_advanced.py

@@ -1279,7 +1279,7 @@ def main(args):
         for name, param in text_encoder_one.named_parameters():
             if "token_embedding" in name:
                 # ensure that dtype is float32, even if rest of the model that isn't trained is loaded in fp16
-                param = param.to(dtype=torch.float32)
+                param.data = param.to(dtype=torch.float32)
                 param.requires_grad = True
                 text_lora_parameters_one.append(param)
             else:
@@ -1288,7 +1288,7 @@ def main(args):
         for name, param in text_encoder_two.named_parameters():
             if "token_embedding" in name:
                 # ensure that dtype is float32, even if rest of the model that isn't trained is loaded in fp16
-                param = param.to(dtype=torch.float32)
+                param.data = param.to(dtype=torch.float32)
                 param.requires_grad = True
                 text_lora_parameters_two.append(param)
             else:
@@ -1725,19 +1725,19 @@ def compute_text_embeddings(prompt, text_encoders, tokenizers):
         num_train_epochs_text_encoder = int(args.train_text_encoder_frac * args.num_train_epochs)
     elif args.train_text_encoder_ti:  # args.train_text_encoder_ti
         num_train_epochs_text_encoder = int(args.train_text_encoder_ti_frac * args.num_train_epochs)
-
+    # flag used for textual inversion
+    pivoted = False
     for epoch in range(first_epoch, args.num_train_epochs):
         # if performing any kind of optimization of text_encoder params
         if args.train_text_encoder or args.train_text_encoder_ti:
             if epoch == num_train_epochs_text_encoder:
                 print("PIVOT HALFWAY", epoch)
                 # stopping optimization of text_encoder params
-                # re setting the optimizer to optimize only on unet params
-                optimizer.param_groups[1]["lr"] = 0.0
-                optimizer.param_groups[2]["lr"] = 0.0
+                # this flag is used to reset the optimizer to optimize only on unet params
+                pivoted = True
 
             else:
-                # still optimizng the text encoder
+                # still optimizing the text encoder
                 text_encoder_one.train()
                 text_encoder_two.train()
                 # set top parameter requires_grad = True for gradient checkpointing works
@@ -1747,6 +1747,12 @@ def compute_text_embeddings(prompt, text_encoders, tokenizers):
 
         unet.train()
         for step, batch in enumerate(train_dataloader):
+            if pivoted:
+                # stopping optimization of text_encoder params
+                # re setting the optimizer to optimize only on unet params
+                optimizer.param_groups[1]["lr"] = 0.0
+                optimizer.param_groups[2]["lr"] = 0.0
+
             with accelerator.accumulate(unet):
                 prompts = batch["prompts"]
                 # encode batch prompts when custom prompts are provided for each image -
@@ -1885,8 +1891,7 @@ def compute_text_embeddings(prompt, text_encoders, tokenizers):
 
                 # every step, we reset the embeddings to the original embeddings.
                 if args.train_text_encoder_ti:
-                    for idx, text_encoder in enumerate(text_encoders):
-                        embedding_handler.retract_embeddings()
+                    embedding_handler.retract_embeddings()
 
             # Checks if the accelerator has performed an optimization step behind the scenes
             if accelerator.sync_gradients: