Make embedding decay work like Adam decay

2 files changed, 9 insertions, 21 deletions

diff --git a/train_ti.py b/train_ti.py
index 0891c49..fc34d27 100644
--- a/train_ti.py
+++ b/train_ti.py
@@ -159,7 +159,7 @@ def parse_args():
     parser.add_argument(
         "--tag_dropout",
         type=float,
-        default=0,
+        default=0.1,
         help="Tag dropout probability.",
     )
     parser.add_argument(
@@ -406,18 +406,12 @@ def parse_args():
         help="Embedding decay target."
     )
     parser.add_argument(
-        "--emb_decay_factor",
-        default=1,
+        "--emb_decay",
+        default=1e-1,
         type=float,
         help="Embedding decay factor."
     )
     parser.add_argument(
-        "--emb_decay_start",
-        default=0,
-        type=float,
-        help="Embedding decay start offset."
-    )
-    parser.add_argument(
         "--noise_timesteps",
         type=int,
         default=1000,
@@ -587,12 +581,10 @@ def main():
         tokenizer=tokenizer,
         sample_scheduler=sample_scheduler,
         checkpoint_output_dir=checkpoint_output_dir,
-        learning_rate=args.learning_rate,
         gradient_checkpointing=args.gradient_checkpointing,
         use_emb_decay=args.use_emb_decay,
         emb_decay_target=args.emb_decay_target,
-        emb_decay_factor=args.emb_decay_factor,
-        emb_decay_start=args.emb_decay_start,
+        emb_decay=args.emb_decay,
         use_ema=args.use_ema,
         ema_inv_gamma=args.ema_inv_gamma,
         ema_power=args.ema_power,
diff --git a/training/strategy/ti.py b/training/strategy/ti.py
index 081180f..eb6730b 100644
--- a/training/strategy/ti.py
+++ b/training/strategy/ti.py
@@ -32,12 +32,10 @@ def textual_inversion_strategy_callbacks(
     seed: int,
     placeholder_tokens: list[str],
     placeholder_token_ids: list[list[int]],
-    learning_rate: float,
     gradient_checkpointing: bool = False,
     use_emb_decay: bool = False,
     emb_decay_target: float = 0.4,
-    emb_decay_factor: float = 1,
-    emb_decay_start: float = 0,
+    emb_decay: float = 1e-2,
     use_ema: bool = False,
     ema_inv_gamma: float = 1.0,
     ema_power: int = 1,
@@ -120,17 +118,15 @@ def textual_inversion_strategy_callbacks(
             yield
 
     def on_after_optimize(lr: float):
-        if ema_embeddings is not None:
-            ema_embeddings.step(text_encoder.text_model.embeddings.temp_token_embedding.parameters())
-
-    @torch.no_grad()
-    def on_after_epoch(lr: float):
         if use_emb_decay:
             text_encoder.text_model.embeddings.normalize(
                 emb_decay_target,
-                min(1.0, max(0.0, emb_decay_factor * ((lr - emb_decay_start) / (learning_rate - emb_decay_start))))
+                min(1.0, emb_decay * lr)
             )
 
+        if ema_embeddings is not None:
+            ema_embeddings.step(text_encoder.text_model.embeddings.temp_token_embedding.parameters())
+
     def on_log():
         if ema_embeddings is not None:
             return {"ema_decay": ema_embeddings.decay}