[Training] make checkpointing compatible when using torch.compile (part II) (#6511)

make checkpointing compatible when using torch.compile.

Author: sayakpaul

examples/dreambooth/train_dreambooth_lora_sdxl.py

@@ -56,6 +56,7 @@
 from diffusers.training_utils import compute_snr
 from diffusers.utils import check_min_version, convert_state_dict_to_diffusers, is_wandb_available
 from diffusers.utils.import_utils import is_xformers_available
+from diffusers.utils.torch_utils import is_compiled_module
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
@@ -1007,6 +1008,11 @@ def main(args):
                 if param.requires_grad:
                     param.data = param.to(torch.float32)
 
+    def unwrap_model(model):
+        model = accelerator.unwrap_model(model)
+        model = model._orig_mod if is_compiled_module(model) else model
+        return model
+
     # create custom saving & loading hooks so that `accelerator.save_state(...)` serializes in a nice format
     def save_model_hook(models, weights, output_dir):
         if accelerator.is_main_process:
@@ -1017,13 +1023,13 @@ def save_model_hook(models, weights, output_dir):
             text_encoder_two_lora_layers_to_save = None
 
             for model in models:
-                if isinstance(model, type(accelerator.unwrap_model(unet))):
+                if isinstance(model, type(unwrap_model(unet))):
                     unet_lora_layers_to_save = convert_state_dict_to_diffusers(get_peft_model_state_dict(model))
-                elif isinstance(model, type(accelerator.unwrap_model(text_encoder_one))):
+                elif isinstance(model, type(unwrap_model(text_encoder_one))):
                     text_encoder_one_lora_layers_to_save = convert_state_dict_to_diffusers(
                         get_peft_model_state_dict(model)
                     )
-                elif isinstance(model, type(accelerator.unwrap_model(text_encoder_two))):
+                elif isinstance(model, type(unwrap_model(text_encoder_two))):
                     text_encoder_two_lora_layers_to_save = convert_state_dict_to_diffusers(
                         get_peft_model_state_dict(model)
                     )
@@ -1048,11 +1054,11 @@ def load_model_hook(models, input_dir):
         while len(models) > 0:
             model = models.pop()
 
-            if isinstance(model, type(accelerator.unwrap_model(unet))):
+            if isinstance(model, type(unwrap_model(unet))):
                 unet_ = model
-            elif isinstance(model, type(accelerator.unwrap_model(text_encoder_one))):
+            elif isinstance(model, type(unwrap_model(text_encoder_one))):
                 text_encoder_one_ = model
-            elif isinstance(model, type(accelerator.unwrap_model(text_encoder_two))):
+            elif isinstance(model, type(unwrap_model(text_encoder_two))):
                 text_encoder_two_ = model
             else:
                 raise ValueError(f"unexpected save model: {model.__class__}")
@@ -1621,16 +1627,16 @@ def compute_text_embeddings(prompt, text_encoders, tokenizers):
     # Save the lora layers
     accelerator.wait_for_everyone()
     if accelerator.is_main_process:
-        unet = accelerator.unwrap_model(unet)
+        unet = unwrap_model(unet)
         unet = unet.to(torch.float32)
         unet_lora_layers = convert_state_dict_to_diffusers(get_peft_model_state_dict(unet))
 
         if args.train_text_encoder:
-            text_encoder_one = accelerator.unwrap_model(text_encoder_one)
+            text_encoder_one = unwrap_model(text_encoder_one)
             text_encoder_lora_layers = convert_state_dict_to_diffusers(
                 get_peft_model_state_dict(text_encoder_one.to(torch.float32))
             )
-            text_encoder_two = accelerator.unwrap_model(text_encoder_two)
+            text_encoder_two = unwrap_model(text_encoder_two)
             text_encoder_2_lora_layers = convert_state_dict_to_diffusers(
                 get_peft_model_state_dict(text_encoder_two.to(torch.float32))
             )