Merge branch 'master' into gradient-clipping

1 files changed, 13 insertions, 13 deletions

diff --git a/modules/sd_models.py b/modules/sd_models.py
index 5075fadb..34c57bfa 100644
--- a/modules/sd_models.py
+++ b/modules/sd_models.py
@@ -163,11 +163,13 @@ def load_model_weights(model, checkpoint_info, vae_file="auto"):
     checkpoint_file = checkpoint_info.filename

     sd_model_hash = checkpoint_info.hash

 

-    vae_file = sd_vae.resolve_vae(checkpoint_file, vae_file=vae_file)

+    if shared.opts.sd_checkpoint_cache > 0 and hasattr(model, "sd_checkpoint_info"):

+        sd_vae.restore_base_vae(model)

+        checkpoints_loaded[model.sd_checkpoint_info] = model.state_dict().copy()

 

-    checkpoint_key = checkpoint_info

+    vae_file = sd_vae.resolve_vae(checkpoint_file, vae_file=vae_file)

 

-    if checkpoint_key not in checkpoints_loaded:

+    if checkpoint_info not in checkpoints_loaded:

         print(f"Loading weights [{sd_model_hash}] from {checkpoint_file}")

 

         pl_sd = torch.load(checkpoint_file, map_location=shared.weight_load_location)

@@ -197,17 +199,15 @@ def load_model_weights(model, checkpoint_info, vae_file="auto"):
 

         model.first_stage_model.to(devices.dtype_vae)

 

-        if shared.opts.sd_checkpoint_cache > 0:

-            # if PR #4035 were to get merged, restore base VAE first before caching

-            checkpoints_loaded[checkpoint_key] = model.state_dict().copy()

-            while len(checkpoints_loaded) > shared.opts.sd_checkpoint_cache:

-                checkpoints_loaded.popitem(last=False)  # LRU

-

     else:

-        vae_name = sd_vae.get_filename(vae_file)

-        print(f"Loading weights [{sd_model_hash}] with {vae_name} VAE from cache")

-        checkpoints_loaded.move_to_end(checkpoint_key)

-        model.load_state_dict(checkpoints_loaded[checkpoint_key])

+        vae_name = sd_vae.get_filename(vae_file) if vae_file else None

+        vae_message = f" with {vae_name} VAE" if vae_name else ""

+        print(f"Loading weights [{sd_model_hash}]{vae_message} from cache")

+        model.load_state_dict(checkpoints_loaded[checkpoint_info])

+

+    if shared.opts.sd_checkpoint_cache > 0:

+        while len(checkpoints_loaded) > shared.opts.sd_checkpoint_cache:

+            checkpoints_loaded.popitem(last=False)  # LRU

 

     model.sd_model_hash = sd_model_hash

     model.sd_model_checkpoint = checkpoint_file