Update

7 files changed, 75 insertions, 51 deletions

diff --git a/data/csv.py b/data/csv.py
index a60733a..d1f3054 100644
--- a/data/csv.py
+++ b/data/csv.py
@@ -1,7 +1,6 @@
 import math
 import torch
 import json
-import numpy as np
 from pathlib import Path
 from PIL import Image
 from torch.utils.data import Dataset, DataLoader, random_split
@@ -9,32 +8,13 @@ from torchvision import transforms
 from typing import Dict, NamedTuple, List, Optional, Union, Callable
 
 from models.clip.prompt import PromptProcessor
+from data.keywords import prompt_to_keywords, keywords_to_prompt
 
 
 def prepare_prompt(prompt: Union[str, Dict[str, str]]):
     return {"content": prompt} if isinstance(prompt, str) else prompt
 
 
-def keywords_to_prompt(prompt: list[str], dropout: float = 0, shuffle: bool = False) -> str:
-    if dropout != 0:
-        prompt = [keyword for keyword in prompt if np.random.random() > dropout]
-    if shuffle:
-        np.random.shuffle(prompt)
-    return ", ".join(prompt)
-
-
-def prompt_to_keywords(prompt: str, expansions: dict[str, str]) -> list[str]:
-    def expand_keyword(keyword: str) -> list[str]:
-        return [keyword] + expansions[keyword].split(", ") if keyword in expansions else [keyword]
-
-    return [
-        kw
-        for keyword in prompt.split(", ")
-        for kw in expand_keyword(keyword)
-        if keyword != ""
-    ]
-
-
 class CSVDataItem(NamedTuple):
     instance_image_path: Path
     class_image_path: Path
diff --git a/data/keywords.py b/data/keywords.py
new file mode 100644
index 0000000..9e656f3
--- /dev/null
+++ b/data/keywords.py
@@ -0,0 +1,21 @@
+import numpy as np
+
+
+def keywords_to_prompt(prompt: list[str], dropout: float = 0, shuffle: bool = False) -> str:
+    if dropout != 0:
+        prompt = [keyword for keyword in prompt if np.random.random() > dropout]
+    if shuffle:
+        np.random.shuffle(prompt)
+    return ", ".join(prompt)
+
+
+def prompt_to_keywords(prompt: str, expansions: dict[str, str] = {}) -> list[str]:
+    def expand_keyword(keyword: str) -> list[str]:
+        return [keyword] + expansions[keyword].split(", ") if keyword in expansions else [keyword]
+
+    return [
+        kw
+        for keyword in prompt.split(", ")
+        for kw in expand_keyword(keyword)
+        if keyword != ""
+    ]
diff --git a/infer.py b/infer.py
index 507d0cf..9c27db4 100644
--- a/infer.py
+++ b/infer.py
@@ -25,6 +25,7 @@ from diffusers import (
 )
 from transformers import CLIPTextModel
 
+from data.keywords import prompt_to_keywords, keywords_to_prompt
 from models.clip.embeddings import patch_managed_embeddings
 from models.clip.tokenizer import MultiCLIPTokenizer
 from pipelines.stable_diffusion.vlpn_stable_diffusion import VlpnStableDiffusion
@@ -49,6 +50,7 @@ default_cmds = {
     "scheduler": "dpmsm",
     "prompt": None,
     "negative_prompt": None,
+    "shuffle": True,
     "image": None,
     "image_noise": .7,
     "width": 768,
@@ -126,6 +128,10 @@ def create_cmd_parser():
         nargs="*",
     )
     parser.add_argument(
+        "--shuffle",
+        type=bool,
+    )
+    parser.add_argument(
         "--image",
         type=str,
     )
@@ -197,7 +203,7 @@ def load_embeddings(pipeline, embeddings_dir):
         pipeline.text_encoder.text_model.embeddings,
         Path(embeddings_dir)
     )
-    print(f"Added {len(added_tokens)} tokens from embeddings dir: {zip(added_tokens, added_ids)}")
+    print(f"Added {len(added_tokens)} tokens from embeddings dir: {list(zip(added_tokens, added_ids))}")
 
 
 def create_pipeline(model, dtype):
@@ -228,20 +234,35 @@ def create_pipeline(model, dtype):
 
 
 @torch.inference_mode()
-def generate(output_dir, pipeline, args):
+def generate(output_dir: Path, pipeline, args):
     if isinstance(args.prompt, str):
         args.prompt = [args.prompt]
 
+    if args.shuffle:
+        args.prompt *= args.batch_size
+        args.batch_size = 1
+        args.prompt = [keywords_to_prompt(prompt_to_keywords(prompt), shuffle=True) for prompt in args.prompt]
+
     now = datetime.datetime.now().strftime("%Y-%m-%dT%H-%M-%S")
-    use_subdirs = len(args.prompt) != 1
-    if use_subdirs:
+    image_dir = []
+
+    if len(args.prompt) != 1:
         if len(args.project) != 0:
             output_dir = output_dir.joinpath(f"{now}_{slugify(args.project)}")
         else:
             output_dir = output_dir.joinpath(now)
+
+        for prompt in args.prompt:
+            dir = output_dir.joinpath(slugify(prompt)[:100])
+            dir.mkdir(parents=True, exist_ok=True)
+            image_dir.append(dir)
+
+            with open(dir.joinpath('prompt.txt'), 'w') as f:
+                f.write(prompt)
     else:
         output_dir = output_dir.joinpath(f"{now}_{slugify(args.prompt[0])[:100]}")
-    output_dir.mkdir(parents=True, exist_ok=True)
+        output_dir.mkdir(parents=True, exist_ok=True)
+        image_dir.append(output_dir)
 
     args.seed = args.seed or torch.random.seed()
 
@@ -293,12 +314,9 @@ def generate(output_dir, pipeline, args):
         ).images
 
         for j, image in enumerate(images):
-            image_dir = output_dir
-            if use_subdirs:
-                image_dir = image_dir.joinpath(slugify(args.prompt[j % len(args.prompt)])[:100])
-            image_dir.mkdir(parents=True, exist_ok=True)
-            image.save(image_dir.joinpath(f"{seed}_{j // len(args.prompt)}.png"))
-            image.save(image_dir.joinpath(f"{seed}_{j // len(args.prompt)}.jpg"), quality=85)
+            dir = image_dir[j % len(args.prompt)]
+            image.save(dir.joinpath(f"{seed}_{j // len(args.prompt)}.png"))
+            image.save(dir.joinpath(f"{seed}_{j // len(args.prompt)}.jpg"), quality=85)
 
     if torch.cuda.is_available():
         torch.cuda.empty_cache()
diff --git a/models/clip/prompt.py b/models/clip/prompt.py
index 9da3955..a7380be 100644
--- a/models/clip/prompt.py
+++ b/models/clip/prompt.py
@@ -1,4 +1,4 @@
-from typing import Union
+from typing import Union, Optional
 
 import torch
 
@@ -16,7 +16,7 @@ class PromptProcessor():
             padding="do_not_pad",
         ).input_ids
 
-    def unify_input_ids(self, input_ids: list[int]):
+    def unify_input_ids(self, input_ids: list[list[int]]):
         return self.tokenizer.pad(
             {"input_ids": input_ids},
             padding=True,
@@ -24,13 +24,15 @@ class PromptProcessor():
             return_tensors="pt"
         )
 
-    def get_embeddings(self, input_ids: torch.IntTensor, attention_mask=None):
+    def get_embeddings(self, input_ids: torch.LongTensor, position_ids: Optional[torch.LongTensor] = None, attention_mask=None):
         prompts = input_ids.shape[0]
 
         input_ids = input_ids.view((-1, self.tokenizer.model_max_length)).to(self.text_encoder.device)
+        if position_ids is not None:
+            position_ids = position_ids.view((-1, self.tokenizer.model_max_length)).to(self.text_encoder.device)
         if attention_mask is not None:
             attention_mask = attention_mask.view((-1, self.tokenizer.model_max_length)).to(self.text_encoder.device)
 
-        text_embeddings = self.text_encoder(input_ids, attention_mask=attention_mask)[0]
+        text_embeddings = self.text_encoder(input_ids, position_ids=position_ids, attention_mask=attention_mask)[0]
         text_embeddings = text_embeddings.view((prompts, -1, text_embeddings.shape[2]))
         return text_embeddings
diff --git a/train_dreambooth.py b/train_dreambooth.py
index 1fd86b1..4d1e0a3 100644
--- a/train_dreambooth.py
+++ b/train_dreambooth.py
@@ -861,13 +861,13 @@ def main():
 
     loop = partial(
         run_model,
-        vae=vae,
-        noise_scheduler=noise_scheduler,
-        unet=unet,
-        prompt_processor=prompt_processor,
-        num_class_images=args.num_class_images,
-        prior_loss_weight=args.prior_loss_weight,
-        seed=args.seed,
+        vae,
+        noise_scheduler,
+        unet,
+        prompt_processor,
+        args.num_class_images,
+        args.prior_loss_weight,
+        args.seed,
     )
 
     # We need to initialize the trackers we use, and also store our configuration.
diff --git a/train_ti.py b/train_ti.py
index 164cf67..98385dd 100644
--- a/train_ti.py
+++ b/train_ti.py
@@ -814,13 +814,13 @@ def main():
 
     loop = partial(
         run_model,
-        vae=vae,
-        noise_scheduler=noise_scheduler,
-        unet=unet,
-        prompt_processor=prompt_processor,
-        num_class_images=args.num_class_images,
-        prior_loss_weight=args.prior_loss_weight,
-        seed=args.seed,
+        vae,
+        noise_scheduler,
+        unet,
+        prompt_processor,
+        args.num_class_images,
+        args.prior_loss_weight,
+        args.seed,
     )
 
     # We need to initialize the trackers we use, and also store our configuration.
diff --git a/training/common.py b/training/common.py
index 99a6e67..ab2741a 100644
--- a/training/common.py
+++ b/training/common.py
@@ -40,7 +40,10 @@ def run_model(
     noisy_latents = noisy_latents.to(dtype=unet.dtype)
 
     # Get the text embedding for conditioning
-    encoder_hidden_states = prompt_processor.get_embeddings(batch["input_ids"], batch["attention_mask"])
+    encoder_hidden_states = prompt_processor.get_embeddings(
+        batch["input_ids"],
+        batch["attention_mask"]
+    )
     encoder_hidden_states = encoder_hidden_states.to(dtype=unet.dtype)
 
     # Predict the noise residual