Upgrade a LoRA rank if the new one is larger so users can increase the rank on an exiting training job and continue training at a higher rank.

jobs/process/BaseSDTrainProcess.py

@@ -1923,15 +1923,23 @@ class BaseSDTrainProcess(BaseTrainProcess):
             for group in optimizer.param_groups:
                 previous_lrs.append(group['lr'])
 
-            try:
-                print_acc(f"Loading optimizer state from {optimizer_state_file_path}")
-                optimizer_state_dict = torch.load(optimizer_state_file_path, weights_only=True)
-                optimizer.load_state_dict(optimizer_state_dict)
-                del optimizer_state_dict
-                flush()
-            except Exception as e:
-                print_acc(f"Failed to load optimizer state from {optimizer_state_file_path}")
-                print_acc(e)
+            load_optimizer = True
+            if self.network is not None:
+                if self.network.did_change_weights:
+                    # do not load optimizer if the network changed, it will result in
+                    # a double state that will oom.
+                    load_optimizer = False
+
+            if load_optimizer:
+                try:
+                    print_acc(f"Loading optimizer state from {optimizer_state_file_path}")
+                    optimizer_state_dict = torch.load(optimizer_state_file_path, weights_only=True)
+                    optimizer.load_state_dict(optimizer_state_dict)
+                    del optimizer_state_dict
+                    flush()
+                except Exception as e:
+                    print_acc(f"Failed to load optimizer state from {optimizer_state_file_path}")
+                    print_acc(e)
 
             # update the optimizer LR from the params
             print_acc(f"Updating optimizer LR from params")

toolkit/network_mixins.py

@@ -46,6 +46,15 @@ ExtractMode = Union[
     'percentage'
 ]
 
+printed_messages = []
+
+
+def print_once(msg):
+    global printed_messages
+    if msg not in printed_messages:
+        print(msg)
+        printed_messages.append(msg)
+
 
 def broadcast_and_multiply(tensor, multiplier):
     # Determine the number of dimensions required
@@ -434,6 +443,8 @@ class ToolkitNetworkMixin:
         self.module_losses: List[torch.Tensor] = []
         self.lorm_train_mode: Literal['local', None] = None
         self.can_merge_in = not is_lorm
+        # will prevent optimizer from loading as it will have double states
+        self.did_change_weights = False
 
     def get_keymap(self: Network, force_weight_mapping=False):
         use_weight_mapping = False
@@ -634,6 +645,46 @@ class ToolkitNetworkMixin:
             if key not in current_state_dict:
                 extra_dict[key] = load_sd[key]
                 to_delete.append(key)
+            elif "lora_down" in key or "lora_up" in key:
+                # handle expanding/shrinking LoRA (linear only)
+                if len(load_sd[key].shape) == 2:
+                    load_value = load_sd[key]                 # from checkpoint
+                    blank_val = current_state_dict[key]       # shape we need in the target model
+                    tgt_h, tgt_w = blank_val.shape
+                    src_h, src_w = load_value.shape
+
+                    if (src_h, src_w) == (tgt_h, tgt_w):
+                        # shapes already match: keep original
+                        pass
+
+                    elif "lora_down" in key and src_h < tgt_h:
+                        print_once(f"Expanding {key} from {load_value.shape} to {blank_val.shape}")
+                        new_val = torch.zeros((tgt_h, tgt_w), device=load_value.device, dtype=load_value.dtype)
+                        new_val[:src_h, :src_w] = load_value  # src_w should already match
+                        load_sd[key] = new_val
+                        self.did_change_weights = True
+
+                    elif "lora_up" in key and src_w < tgt_w:
+                        print_once(f"Expanding {key} from {load_value.shape} to {blank_val.shape}")
+                        new_val = torch.zeros((tgt_h, tgt_w), device=load_value.device, dtype=load_value.dtype)
+                        new_val[:src_h, :src_w] = load_value  # src_h should already match
+                        load_sd[key] = new_val
+                        self.did_change_weights = True
+
+                    elif "lora_down" in key and src_h > tgt_h:
+                        print_once(f"Shrinking {key} from {load_value.shape} to {blank_val.shape}")
+                        load_sd[key] = load_value[:tgt_h, :tgt_w]
+                        self.did_change_weights = True
+
+                    elif "lora_up" in key and src_w > tgt_w:
+                        print_once(f"Shrinking {key} from {load_value.shape} to {blank_val.shape}")
+                        load_sd[key] = load_value[:tgt_h, :tgt_w]
+                        self.did_change_weights = True
+
+                    else:
+                        # unexpected mismatch (e.g., both dims differ in a way that doesn't match lora_up/down semantics)
+                        raise ValueError(f"Unhandled LoRA shape change for {key}: src={load_value.shape}, tgt={blank_val.shape}")
+
         for key in to_delete:
             del load_sd[key]