Fixed accuracy calc, other improvements

5 files changed, 74 insertions, 60 deletions

diff --git a/data/csv.py b/data/csv.py
index 803271b..af36d9e 100644
--- a/data/csv.py
+++ b/data/csv.py
@@ -151,7 +151,7 @@ class CSVDataModule():
 
         num_images = len(items)
 
-        valid_set_size = self.valid_set_size if self.valid_set_size is not None else int(num_images * 0.1)
+        valid_set_size = self.valid_set_size if self.valid_set_size is not None else int(num_images * 0.2)
         valid_set_size = max(valid_set_size, 1)
         train_set_size = num_images - valid_set_size
 
diff --git a/models/clip/tokenizer.py b/models/clip/tokenizer.py
index a3e6e70..37d69a9 100644
--- a/models/clip/tokenizer.py
+++ b/models/clip/tokenizer.py
@@ -15,6 +15,10 @@ class MultiCLIPTokenizer(CLIPTokenizer):
     def __init__(self, *args, **kwargs):
         super().__init__(*args, **kwargs)
         self.token_map: dict[int, list[int]] = {}
+        self.vector_shuffle = False
+
+    def set_use_vector_shuffle(self, enable: bool):
+        self.vector_shuffle = enable
 
     def add_multi_tokens(self, new_tokens: Union[str, list[str]], num_vectors: Union[int, list[int]] = 1) -> MultiCLIPTokenizerItem:
         if isinstance(new_tokens, list):
@@ -42,11 +46,11 @@ class MultiCLIPTokenizer(CLIPTokenizer):
 
         return MultiCLIPTokenizerItem(new_tokens, ids)
 
-    def expand_id(self, id: int, vector_shuffle=True):
+    def expand_id(self, id: int):
         if id in self.token_map:
             tokens = self.token_map[id]
 
-            if vector_shuffle and len(tokens) > 2:
+            if self.vector_shuffle and len(tokens) > 2:
                 subtokens = tokens[1:-1]
                 np.random.shuffle(subtokens)
                 tokens = tokens[:1] + subtokens + tokens[-1:]
@@ -55,21 +59,21 @@ class MultiCLIPTokenizer(CLIPTokenizer):
         else:
             return [id]
 
-    def expand_ids(self, ids: list[int], vector_shuffle=True):
+    def expand_ids(self, ids: list[int]):
         return [
             new_id
             for id in ids
-            for new_id in self.expand_id(id, vector_shuffle)
+            for new_id in self.expand_id(id)
         ]
 
-    def _call_one(self, text, *args, vector_shuffle=True, **kwargs):
+    def _call_one(self, text, *args, **kwargs):
         result = super()._call_one(text, *args, **kwargs)
 
         is_batched = isinstance(result.input_ids, (list, tuple)) and isinstance(result.input_ids[0], list)
 
         if is_batched:
-            result.input_ids = [self.expand_ids(batch, vector_shuffle) for batch in result.input_ids]
+            result.input_ids = [self.expand_ids(batch) for batch in result.input_ids]
         else:
-            result.input_ids = self.expand_ids(result.input_ids, vector_shuffle)
+            result.input_ids = self.expand_ids(result.input_ids)
 
         return result
diff --git a/train_dreambooth.py b/train_dreambooth.py
index 8fd78f1..1ebcfe3 100644
--- a/train_dreambooth.py
+++ b/train_dreambooth.py
@@ -232,6 +232,30 @@ def parse_args():
         help="Number of restart cycles in the lr scheduler (if supported)."
     )
     parser.add_argument(
+        "--lr_warmup_func",
+        type=str,
+        default="cos",
+        help='Choose between ["linear", "cos"]'
+    )
+    parser.add_argument(
+        "--lr_warmup_exp",
+        type=int,
+        default=1,
+        help='If lr_warmup_func is "cos", exponent to modify the function'
+    )
+    parser.add_argument(
+        "--lr_annealing_func",
+        type=str,
+        default="cos",
+        help='Choose between ["linear", "half_cos", "cos"]'
+    )
+    parser.add_argument(
+        "--lr_annealing_exp",
+        type=int,
+        default=3,
+        help='If lr_annealing_func is "half_cos" or "cos", exponent to modify the function'
+    )
+    parser.add_argument(
         "--use_ema",
         action="store_true",
         default=True,
@@ -760,6 +784,10 @@ def main():
         lr_scheduler = get_one_cycle_schedule(
             optimizer=optimizer,
             num_training_steps=args.max_train_steps * args.gradient_accumulation_steps,
+            warmup=args.lr_warmup_func,
+            annealing=args.lr_annealing_func,
+            warmup_exp=args.lr_warmup_exp,
+            annealing_exp=args.lr_annealing_exp,
         )
     elif args.lr_scheduler == "cosine_with_restarts":
         lr_scheduler = get_cosine_with_hard_restarts_schedule_with_warmup(
@@ -913,7 +941,7 @@ def main():
         else:
             loss = F.mse_loss(model_pred.float(), target.float(), reduction="mean")
 
-        acc = (model_pred == latents).float().mean()
+        acc = (model_pred == target).float().mean()
 
         return loss, acc, bsz
 
diff --git a/train_ti.py b/train_ti.py
index 19348e5..20a3190 100644
--- a/train_ti.py
+++ b/train_ti.py
@@ -225,6 +225,30 @@ def parse_args():
         help="Number of restart cycles in the lr scheduler."
     )
     parser.add_argument(
+        "--lr_warmup_func",
+        type=str,
+        default="cos",
+        help='Choose between ["linear", "cos"]'
+    )
+    parser.add_argument(
+        "--lr_warmup_exp",
+        type=int,
+        default=1,
+        help='If lr_warmup_func is "cos", exponent to modify the function'
+    )
+    parser.add_argument(
+        "--lr_annealing_func",
+        type=str,
+        default="cos",
+        help='Choose between ["linear", "half_cos", "cos"]'
+    )
+    parser.add_argument(
+        "--lr_annealing_exp",
+        type=int,
+        default=2,
+        help='If lr_annealing_func is "half_cos" or "cos", exponent to modify the function'
+    )
+    parser.add_argument(
         "--use_8bit_adam",
         action="store_true",
         help="Whether or not to use 8-bit Adam from bitsandbytes."
@@ -510,6 +534,8 @@ def main():
     checkpoint_scheduler = DPMSolverMultistepScheduler.from_pretrained(
         args.pretrained_model_name_or_path, subfolder='scheduler')
 
+    tokenizer.set_use_vector_shuffle(True)
+
     vae.enable_slicing()
     vae.set_use_memory_efficient_attention_xformers(True)
     unet.set_use_memory_efficient_attention_xformers(True)
@@ -559,7 +585,7 @@ def main():
         )
 
     if args.find_lr:
-        args.learning_rate = 1e2
+        args.learning_rate = 1e3
 
     # Use 8-bit Adam for lower memory usage or to fine-tune the model in 16GB GPUs
     if args.use_8bit_adam:
@@ -706,6 +732,10 @@ def main():
         lr_scheduler = get_one_cycle_schedule(
             optimizer=optimizer,
             num_training_steps=args.max_train_steps * args.gradient_accumulation_steps,
+            warmup=args.lr_warmup_func,
+            annealing=args.lr_annealing_func,
+            warmup_exp=args.lr_warmup_exp,
+            annealing_exp=args.lr_annealing_exp,
         )
     elif args.lr_scheduler == "cosine_with_restarts":
         lr_scheduler = get_cosine_with_hard_restarts_schedule_with_warmup(
@@ -796,13 +826,13 @@ def main():
         else:
             loss = F.mse_loss(model_pred.float(), target.float(), reduction="mean")
 
-        acc = (model_pred == latents).float().mean()
+        acc = (model_pred == target).float().mean()
 
         return loss, acc, bsz
 
     if args.find_lr:
         lr_finder = LRFinder(accelerator, text_encoder, optimizer, train_dataloader, val_dataloader, loop)
-        lr_finder.run(min_lr=1e-6, num_train_batches=1)
+        lr_finder.run(min_lr=1e-4)
 
         plt.savefig(basepath.joinpath("lr.png"))
         plt.close()
diff --git a/training/ti.py b/training/ti.py
deleted file mode 100644
index 031fe48..0000000
--- a/training/ti.py
+++ /dev/null
@@ -1,48 +0,0 @@
-from typing import Optional
-
-import torch
-import torch.nn as nn
-
-from transformers.models.clip import CLIPTextModel, CLIPTextConfig
-from transformers.models.clip.modeling_clip import CLIPTextEmbeddings
-
-
-def patch_trainable_embeddings(text_encoder: CLIPTextModel, new_ids: list[int]):
-    text_embeddings = TrainableEmbeddings(text_encoder.config, text_encoder.text_model.embeddings, new_ids)
-    text_encoder.text_model.embeddings = text_embeddings
-
-
-class TrainableEmbeddings(CLIPTextEmbeddings):
-    def __init__(self, config: CLIPTextConfig, embeddings: CLIPTextEmbeddings, new_ids: list[int]):
-        super().__init__(config)
-
-        self.token_embedding = embeddings.token_embedding
-        self.position_embedding = embeddings.position_embedding
-
-        self.train_indices = torch.tensor(new_ids)
-
-        self.trainable_embedding = nn.Embedding(self.token_embedding.num_embeddings, self.token_embedding.embedding_dim)
-        self.trainable_embedding.weight.data.zero_()
-        self.trainable_embedding.weight.data[self.train_indices] = self.token_embedding.weight.data[self.train_indices]
-
-    def forward(
-        self,
-        input_ids: Optional[torch.LongTensor] = None,
-        position_ids: Optional[torch.LongTensor] = None,
-        inputs_embeds: Optional[torch.FloatTensor] = None,
-    ) -> torch.Tensor:
-        device = input_ids.device
-        seq_length = input_ids.shape[-1] if input_ids is not None else inputs_embeds.shape[-2]
-
-        if position_ids is None:
-            position_ids = self.position_ids[:, :seq_length]
-
-        if inputs_embeds is None:
-            mask = torch.isin(input_ids, self.train_indices.to(device))
-            inputs_embeds = self.token_embedding(input_ids)
-            inputs_embeds[mask] = self.trainable_embedding(input_ids)[mask]
-
-        position_embeddings = self.position_embedding(position_ids)
-        embeddings = inputs_embeds + position_embeddings
-
-        return embeddings