From 820f1dc96b1979d7e92170c161db281ee8bd988b Mon Sep 17 00:00:00 2001
From: AUTOMATIC <16777216c@gmail.com>
Date: Sun, 2 Oct 2022 15:03:39 +0300
Subject: initial support for training textual inversion

---
 modules/textual_inversion/dataset.py | 76 ++++++++++++++++++++++++++++++++++++
 1 file changed, 76 insertions(+)
 create mode 100644 modules/textual_inversion/dataset.py

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
new file mode 100644
index 00000000..7e134a08
--- /dev/null
+++ b/modules/textual_inversion/dataset.py
@@ -0,0 +1,76 @@
+import os
+import numpy as np
+import PIL
+import torch
+from PIL import Image
+from torch.utils.data import Dataset
+from torchvision import transforms
+
+import random
+import tqdm
+
+
+class PersonalizedBase(Dataset):
+    def __init__(self, data_root, size=None, repeats=100, flip_p=0.5, placeholder_token="*", width=512, height=512, model=None, device=None, template_file=None):
+
+        self.placeholder_token = placeholder_token
+
+        self.size = size
+        self.width = width
+        self.height = height
+        self.flip = transforms.RandomHorizontalFlip(p=flip_p)
+
+        self.dataset = []
+
+        with open(template_file, "r") as file:
+            lines = [x.strip() for x in file.readlines()]
+
+        self.lines = lines
+
+        assert data_root, 'dataset directory not specified'
+
+        self.image_paths = [os.path.join(data_root, file_path) for file_path in os.listdir(data_root)]
+        print("Preparing dataset...")
+        for path in tqdm.tqdm(self.image_paths):
+            image = Image.open(path)
+            image = image.convert('RGB')
+            image = image.resize((self.width, self.height), PIL.Image.BICUBIC)
+
+            filename = os.path.basename(path)
+            filename_tokens = os.path.splitext(filename)[0].replace('_', '-').replace(' ', '-').split('-')
+            filename_tokens = [token for token in filename_tokens if token.isalpha()]
+
+            npimage = np.array(image).astype(np.uint8)
+            npimage = (npimage / 127.5 - 1.0).astype(np.float32)
+
+            torchdata = torch.from_numpy(npimage).to(device=device, dtype=torch.float32)
+            torchdata = torch.moveaxis(torchdata, 2, 0)
+
+            init_latent = model.get_first_stage_encoding(model.encode_first_stage(torchdata.unsqueeze(dim=0))).squeeze()
+
+            self.dataset.append((init_latent, filename_tokens))
+
+        self.length = len(self.dataset) * repeats
+
+        self.initial_indexes = np.arange(self.length) % len(self.dataset)
+        self.indexes = None
+        self.shuffle()
+
+    def shuffle(self):
+        self.indexes = self.initial_indexes[torch.randperm(self.initial_indexes.shape[0])]
+
+    def __len__(self):
+        return self.length
+
+    def __getitem__(self, i):
+        if i % len(self.dataset) == 0:
+            self.shuffle()
+
+        index = self.indexes[i % len(self.indexes)]
+        x, filename_tokens = self.dataset[index]
+
+        text = random.choice(self.lines)
+        text = text.replace("[name]", self.placeholder_token)
+        text = text.replace("[filewords]", ' '.join(filename_tokens))
+
+        return x, text
-- 
cgit v1.2.3


From 6785331e22d6a488fbf5905fab56d7fec867e038 Mon Sep 17 00:00:00 2001
From: AUTOMATIC <16777216c@gmail.com>
Date: Sun, 2 Oct 2022 22:59:01 +0300
Subject: keep textual inversion dataset latents in CPU memory to save a bit of
 VRAM

---
 modules/textual_inversion/dataset.py           | 2 ++
 modules/textual_inversion/textual_inversion.py | 3 +++
 modules/ui.py                                  | 4 ++--
 3 files changed, 7 insertions(+), 2 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index 7e134a08..e8394ff6 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -8,6 +8,7 @@ from torchvision import transforms
 
 import random
 import tqdm
+from modules import devices
 
 
 class PersonalizedBase(Dataset):
@@ -47,6 +48,7 @@ class PersonalizedBase(Dataset):
             torchdata = torch.moveaxis(torchdata, 2, 0)
 
             init_latent = model.get_first_stage_encoding(model.encode_first_stage(torchdata.unsqueeze(dim=0))).squeeze()
+            init_latent = init_latent.to(devices.cpu)
 
             self.dataset.append((init_latent, filename_tokens))
 
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index d4e250d8..8686f534 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -212,7 +212,10 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, steps,
 
         with torch.autocast("cuda"):
             c = cond_model([text])
+
+            x = x.to(devices.device)
             loss = shared.sd_model(x.unsqueeze(0), c)[0]
+            del x
 
             losses[embedding.step % losses.shape[0]] = loss.item()
 
diff --git a/modules/ui.py b/modules/ui.py
index e7bde53b..d9d02ece 100644
--- a/modules/ui.py
+++ b/modules/ui.py
@@ -1002,8 +1002,8 @@ def create_ui(wrap_gradio_gpu_call):
                     log_directory = gr.Textbox(label='Log directory', placeholder="Path to directory where to write outputs", value="textual_inversion")
                     template_file = gr.Textbox(label='Prompt template file', value=os.path.join(script_path, "textual_inversion_templates", "style_filewords.txt"))
                     steps = gr.Number(label='Max steps', value=100000, precision=0)
-                    create_image_every = gr.Number(label='Save an image to log directory every N steps, 0 to disable', value=1000, precision=0)
-                    save_embedding_every = gr.Number(label='Save a copy of embedding to log directory every N steps, 0 to disable', value=1000, precision=0)
+                    create_image_every = gr.Number(label='Save an image to log directory every N steps, 0 to disable', value=500, precision=0)
+                    save_embedding_every = gr.Number(label='Save a copy of embedding to log directory every N steps, 0 to disable', value=500, precision=0)
 
                     with gr.Row():
                         with gr.Column(scale=2):
-- 
cgit v1.2.3


From 5ef0baf5eaec7f21a1666af424405cbee19f3764 Mon Sep 17 00:00:00 2001
From: AUTOMATIC <16777216c@gmail.com>
Date: Tue, 4 Oct 2022 08:52:11 +0300
Subject: add support for gelbooru tags in filenames for textual inversion

---
 modules/textual_inversion/dataset.py    | 7 +++++--
 modules/textual_inversion/preprocess.py | 4 +++-
 2 files changed, 8 insertions(+), 3 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index e8394ff6..7c44ea5b 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -9,6 +9,9 @@ from torchvision import transforms
 import random
 import tqdm
 from modules import devices
+import re
+
+re_tag = re.compile(r"[a-zA-Z][_\w\d()]+")
 
 
 class PersonalizedBase(Dataset):
@@ -38,8 +41,8 @@ class PersonalizedBase(Dataset):
             image = image.resize((self.width, self.height), PIL.Image.BICUBIC)
 
             filename = os.path.basename(path)
-            filename_tokens = os.path.splitext(filename)[0].replace('_', '-').replace(' ', '-').split('-')
-            filename_tokens = [token for token in filename_tokens if token.isalpha()]
+            filename_tokens = os.path.splitext(filename)[0]
+            filename_tokens = re_tag.findall(filename_tokens)
 
             npimage = np.array(image).astype(np.uint8)
             npimage = (npimage / 127.5 - 1.0).astype(np.float32)
diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py
index 209e928f..f545a993 100644
--- a/modules/textual_inversion/preprocess.py
+++ b/modules/textual_inversion/preprocess.py
@@ -26,7 +26,9 @@ def preprocess(process_src, process_dst, process_flip, process_split, process_ca
         if process_caption:
             caption = "-" + shared.interrogator.generate_caption(image)
         else:
-            caption = ""
+            caption = filename
+            caption = os.path.splitext(caption)[0]
+            caption = os.path.basename(caption)
 
         image.save(os.path.join(dst, f"{index:05}-{subindex[0]}{caption}.png"))
         subindex[0] += 1
-- 
cgit v1.2.3


From 3110f895b2718a3a25aae419fdf5c87c177ec9f4 Mon Sep 17 00:00:00 2001
From: alg-wiki <alg.4chan@gmail.com>
Date: Mon, 10 Oct 2022 17:07:46 +0900
Subject: Textual Inversion: Added custom training image size and number of
 repeats per input image in a single epoch

---
 modules/textual_inversion/dataset.py           |  6 +++---
 modules/textual_inversion/preprocess.py        |  4 ++--
 modules/textual_inversion/textual_inversion.py | 15 ++++++++++++---
 modules/ui.py                                  |  8 +++++++-
 4 files changed, 24 insertions(+), 9 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index 7c44ea5b..acc4ce59 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -15,13 +15,13 @@ re_tag = re.compile(r"[a-zA-Z][_\w\d()]+")
 
 
 class PersonalizedBase(Dataset):
-    def __init__(self, data_root, size=None, repeats=100, flip_p=0.5, placeholder_token="*", width=512, height=512, model=None, device=None, template_file=None):
+    def __init__(self, data_root, size, repeats, flip_p=0.5, placeholder_token="*", model=None, device=None, template_file=None):
 
         self.placeholder_token = placeholder_token
 
         self.size = size
-        self.width = width
-        self.height = height
+        self.width = size
+        self.height = size
         self.flip = transforms.RandomHorizontalFlip(p=flip_p)
 
         self.dataset = []
diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py
index f1c002a2..b3de6fd7 100644
--- a/modules/textual_inversion/preprocess.py
+++ b/modules/textual_inversion/preprocess.py
@@ -7,8 +7,8 @@ import tqdm
 from modules import shared, images
 
 
-def preprocess(process_src, process_dst, process_flip, process_split, process_caption):
-    size = 512
+def preprocess(process_src, process_dst, process_size, process_flip, process_split, process_caption):
+    size = process_size
     src = os.path.abspath(process_src)
     dst = os.path.abspath(process_dst)
 
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index cd9f3498..e34dc2e8 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -6,6 +6,7 @@ import torch
 import tqdm
 import html
 import datetime
+import math
 
 
 from modules import shared, devices, sd_hijack, processing, sd_models
@@ -156,7 +157,7 @@ def create_embedding(name, num_vectors_per_token, init_text='*'):
     return fn
 
 
-def train_embedding(embedding_name, learn_rate, data_root, log_directory, steps, create_image_every, save_embedding_every, template_file):
+def train_embedding(embedding_name, learn_rate, data_root, log_directory, training_size, steps, num_repeats, create_image_every, save_embedding_every, template_file):
     assert embedding_name, 'embedding not selected'
 
     shared.state.textinfo = "Initializing textual inversion training..."
@@ -182,7 +183,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, steps,
 
     shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..."
     with torch.autocast("cuda"):
-        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, size=512, placeholder_token=embedding_name, model=shared.sd_model, device=devices.device, template_file=template_file)
+        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, size=training_size, repeats=num_repeats, placeholder_token=embedding_name, model=shared.sd_model, device=devices.device, template_file=template_file)
 
     hijack = sd_hijack.model_hijack
 
@@ -200,6 +201,9 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, steps,
     if ititial_step > steps:
         return embedding, filename
 
+    tr_img_len = len([os.path.join(data_root, file_path) for file_path in os.listdir(data_root)])
+    epoch_len = (tr_img_len * num_repeats) + tr_img_len
+
     pbar = tqdm.tqdm(enumerate(ds), total=steps-ititial_step)
     for i, (x, text) in pbar:
         embedding.step = i + ititial_step
@@ -223,7 +227,10 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, steps,
             loss.backward()
             optimizer.step()
 
-        pbar.set_description(f"loss: {losses.mean():.7f}")
+        epoch_num = math.floor(embedding.step / epoch_len)
+        epoch_step = embedding.step - (epoch_num * epoch_len)
+
+        pbar.set_description(f"[Epoch {epoch_num}: {epoch_step}/{epoch_len}]loss: {losses.mean():.7f}")
 
         if embedding.step > 0 and embedding_dir is not None and embedding.step % save_embedding_every == 0:
             last_saved_file = os.path.join(embedding_dir, f'{embedding_name}-{embedding.step}.pt')
@@ -236,6 +243,8 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, steps,
                 sd_model=shared.sd_model,
                 prompt=text,
                 steps=20,
+                height=training_size,
+                width=training_size,
                 do_not_save_grid=True,
                 do_not_save_samples=True,
             )
diff --git a/modules/ui.py b/modules/ui.py
index 2231a8ed..f821fd8d 100644
--- a/modules/ui.py
+++ b/modules/ui.py
@@ -1029,6 +1029,7 @@ def create_ui(wrap_gradio_gpu_call):
 
                     process_src = gr.Textbox(label='Source directory')
                     process_dst = gr.Textbox(label='Destination directory')
+                    process_size = gr.Slider(minimum=64, maximum=2048, step=64, label="Size (width and height)", value=512)
 
                     with gr.Row():
                         process_flip = gr.Checkbox(label='Create flipped copies')
@@ -1043,13 +1044,15 @@ def create_ui(wrap_gradio_gpu_call):
                             run_preprocess = gr.Button(value="Preprocess", variant='primary')
 
                 with gr.Group():
-                    gr.HTML(value="<p style='margin-bottom: 0.7em'>Train an embedding; must specify a directory with a set of 512x512 images</p>")
+                    gr.HTML(value="<p style='margin-bottom: 0.7em'>Train an embedding; must specify a directory with a set of 1:1 ratio images</p>")
                     train_embedding_name = gr.Dropdown(label='Embedding', choices=sorted(sd_hijack.model_hijack.embedding_db.word_embeddings.keys()))
                     learn_rate = gr.Number(label='Learning rate', value=5.0e-03)
                     dataset_directory = gr.Textbox(label='Dataset directory', placeholder="Path to directory with input images")
                     log_directory = gr.Textbox(label='Log directory', placeholder="Path to directory where to write outputs", value="textual_inversion")
                     template_file = gr.Textbox(label='Prompt template file', value=os.path.join(script_path, "textual_inversion_templates", "style_filewords.txt"))
+                    training_size = gr.Slider(minimum=64, maximum=2048, step=64, label="Size (width and height)", value=512)
                     steps = gr.Number(label='Max steps', value=100000, precision=0)
+                    num_repeats = gr.Number(label='Number of repeats for a single input image per epoch', value=100, precision=0)
                     create_image_every = gr.Number(label='Save an image to log directory every N steps, 0 to disable', value=500, precision=0)
                     save_embedding_every = gr.Number(label='Save a copy of embedding to log directory every N steps, 0 to disable', value=500, precision=0)
 
@@ -1092,6 +1095,7 @@ def create_ui(wrap_gradio_gpu_call):
             inputs=[
                 process_src,
                 process_dst,
+                process_size,
                 process_flip,
                 process_split,
                 process_caption,
@@ -1110,7 +1114,9 @@ def create_ui(wrap_gradio_gpu_call):
                 learn_rate,
                 dataset_directory,
                 log_directory,
+                training_size,
                 steps,
+                num_repeats,
                 create_image_every,
                 save_embedding_every,
                 template_file,
-- 
cgit v1.2.3


From 04c745ea4f81518999927fee5f78500560c25e29 Mon Sep 17 00:00:00 2001
From: alg-wiki <alg.4chan@gmail.com>
Date: Mon, 10 Oct 2022 22:35:35 +0900
Subject: Custom Width and Height

---
 modules/textual_inversion/dataset.py           |  7 +++----
 modules/textual_inversion/preprocess.py        | 19 ++++++++++---------
 modules/textual_inversion/textual_inversion.py | 11 +++++------
 modules/ui.py                                  | 12 ++++++++----
 4 files changed, 26 insertions(+), 23 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index acc4ce59..bcf772d2 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -15,13 +15,12 @@ re_tag = re.compile(r"[a-zA-Z][_\w\d()]+")
 
 
 class PersonalizedBase(Dataset):
-    def __init__(self, data_root, size, repeats, flip_p=0.5, placeholder_token="*", model=None, device=None, template_file=None):
+    def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", model=None, device=None, template_file=None):
 
         self.placeholder_token = placeholder_token
 
-        self.size = size
-        self.width = size
-        self.height = size
+        self.width = width
+        self.height = height
         self.flip = transforms.RandomHorizontalFlip(p=flip_p)
 
         self.dataset = []
diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py
index b3de6fd7..d7efdef2 100644
--- a/modules/textual_inversion/preprocess.py
+++ b/modules/textual_inversion/preprocess.py
@@ -7,8 +7,9 @@ import tqdm
 from modules import shared, images
 
 
-def preprocess(process_src, process_dst, process_size, process_flip, process_split, process_caption):
-    size = process_size
+def preprocess(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption):
+    width = process_width
+    height = process_height
     src = os.path.abspath(process_src)
     dst = os.path.abspath(process_dst)
 
@@ -55,23 +56,23 @@ def preprocess(process_src, process_dst, process_size, process_flip, process_spl
         is_wide = ratio < 1 / 1.35
 
         if process_split and is_tall:
-            img = img.resize((size, size * img.height // img.width))
+            img = img.resize((width, height * img.height // img.width))
 
-            top = img.crop((0, 0, size, size))
+            top = img.crop((0, 0, width, height))
             save_pic(top, index)
 
-            bot = img.crop((0, img.height - size, size, img.height))
+            bot = img.crop((0, img.height - height, width, img.height))
             save_pic(bot, index)
         elif process_split and is_wide:
-            img = img.resize((size * img.width // img.height, size))
+            img = img.resize((width * img.width // img.height, height))
 
-            left = img.crop((0, 0, size, size))
+            left = img.crop((0, 0, width, height))
             save_pic(left, index)
 
-            right = img.crop((img.width - size, 0, img.width, size))
+            right = img.crop((img.width - width, 0, img.width, height))
             save_pic(right, index)
         else:
-            img = images.resize_image(1, img, size, size)
+            img = images.resize_image(1, img, width, height)
             save_pic(img, index)
 
         shared.state.nextjob()
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index 769682ea..5965c5a0 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -6,7 +6,6 @@ import torch
 import tqdm
 import html
 import datetime
-import math
 
 
 from modules import shared, devices, sd_hijack, processing, sd_models
@@ -157,7 +156,7 @@ def create_embedding(name, num_vectors_per_token, init_text='*'):
     return fn
 
 
-def train_embedding(embedding_name, learn_rate, data_root, log_directory, training_size, steps, num_repeats, create_image_every, save_embedding_every, template_file):
+def train_embedding(embedding_name, learn_rate, data_root, log_directory, training_width, training_height, steps, num_repeats, create_image_every, save_embedding_every, template_file):
     assert embedding_name, 'embedding not selected'
 
     shared.state.textinfo = "Initializing textual inversion training..."
@@ -183,7 +182,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
 
     shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..."
     with torch.autocast("cuda"):
-        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, size=training_size, repeats=num_repeats, placeholder_token=embedding_name, model=shared.sd_model, device=devices.device, template_file=template_file)
+        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=num_repeats, placeholder_token=embedding_name, model=shared.sd_model, device=devices.device, template_file=template_file)
 
     hijack = sd_hijack.model_hijack
 
@@ -227,7 +226,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
             loss.backward()
             optimizer.step()
 
-        epoch_num = math.floor(embedding.step / epoch_len)
+        epoch_num = embedding.step // epoch_len
         epoch_step = embedding.step - (epoch_num * epoch_len) + 1
 
         pbar.set_description(f"[Epoch {epoch_num}: {epoch_step}/{epoch_len}]loss: {losses.mean():.7f}")
@@ -243,8 +242,8 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
                 sd_model=shared.sd_model,
                 prompt=text,
                 steps=20,
-                height=training_size,
-                width=training_size,
+				height=training_height,
+				width=training_width,
                 do_not_save_grid=True,
                 do_not_save_samples=True,
             )
diff --git a/modules/ui.py b/modules/ui.py
index f821fd8d..8c06ad7c 100644
--- a/modules/ui.py
+++ b/modules/ui.py
@@ -1029,7 +1029,8 @@ def create_ui(wrap_gradio_gpu_call):
 
                     process_src = gr.Textbox(label='Source directory')
                     process_dst = gr.Textbox(label='Destination directory')
-                    process_size = gr.Slider(minimum=64, maximum=2048, step=64, label="Size (width and height)", value=512)
+                    process_width = gr.Slider(minimum=64, maximum=2048, step=64, label="Width", value=512)
+                    process_height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height", value=512)
 
                     with gr.Row():
                         process_flip = gr.Checkbox(label='Create flipped copies')
@@ -1050,7 +1051,8 @@ def create_ui(wrap_gradio_gpu_call):
                     dataset_directory = gr.Textbox(label='Dataset directory', placeholder="Path to directory with input images")
                     log_directory = gr.Textbox(label='Log directory', placeholder="Path to directory where to write outputs", value="textual_inversion")
                     template_file = gr.Textbox(label='Prompt template file', value=os.path.join(script_path, "textual_inversion_templates", "style_filewords.txt"))
-                    training_size = gr.Slider(minimum=64, maximum=2048, step=64, label="Size (width and height)", value=512)
+                    training_width = gr.Slider(minimum=64, maximum=2048, step=64, label="Width", value=512)
+                    training_height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height", value=512)
                     steps = gr.Number(label='Max steps', value=100000, precision=0)
                     num_repeats = gr.Number(label='Number of repeats for a single input image per epoch', value=100, precision=0)
                     create_image_every = gr.Number(label='Save an image to log directory every N steps, 0 to disable', value=500, precision=0)
@@ -1095,7 +1097,8 @@ def create_ui(wrap_gradio_gpu_call):
             inputs=[
                 process_src,
                 process_dst,
-                process_size,
+                process_width,
+                process_height,
                 process_flip,
                 process_split,
                 process_caption,
@@ -1114,7 +1117,8 @@ def create_ui(wrap_gradio_gpu_call):
                 learn_rate,
                 dataset_directory,
                 log_directory,
-                training_size,
+                training_width,
+                training_height,
                 steps,
                 num_repeats,
                 create_image_every,
-- 
cgit v1.2.3


From bc3e183b739913e7be91213a256f038b10eb71e9 Mon Sep 17 00:00:00 2001
From: alg-wiki <alg.4chan@gmail.com>
Date: Tue, 11 Oct 2022 04:30:13 +0900
Subject: Textual Inversion: Preprocess and Training will only pick-up image
 files

---
 modules/textual_inversion/dataset.py           | 3 ++-
 modules/textual_inversion/preprocess.py        | 3 ++-
 modules/textual_inversion/textual_inversion.py | 3 ++-
 3 files changed, 6 insertions(+), 3 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index bcf772d2..d4baf066 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -22,6 +22,7 @@ class PersonalizedBase(Dataset):
         self.width = width
         self.height = height
         self.flip = transforms.RandomHorizontalFlip(p=flip_p)
+        self.extns = [".jpg",".jpeg",".png"]
 
         self.dataset = []
 
@@ -32,7 +33,7 @@ class PersonalizedBase(Dataset):
 
         assert data_root, 'dataset directory not specified'
 
-        self.image_paths = [os.path.join(data_root, file_path) for file_path in os.listdir(data_root)]
+        self.image_paths = [os.path.join(data_root, file_path) for file_path in os.listdir(data_root) if os.path.splitext(file_path.casefold())[1] in self.extns]
         print("Preparing dataset...")
         for path in tqdm.tqdm(self.image_paths):
             image = Image.open(path)
diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py
index d7efdef2..b6c78cf8 100644
--- a/modules/textual_inversion/preprocess.py
+++ b/modules/textual_inversion/preprocess.py
@@ -12,12 +12,13 @@ def preprocess(process_src, process_dst, process_width, process_height, process_
     height = process_height
     src = os.path.abspath(process_src)
     dst = os.path.abspath(process_dst)
+    extns = [".jpg",".jpeg",".png"]
 
     assert src != dst, 'same directory specified as source and destination'
 
     os.makedirs(dst, exist_ok=True)
 
-    files = os.listdir(src)
+    files = [i for i in os.listdir(src) if os.path.splitext(i.casefold())[1] in extns]
 
     shared.state.textinfo = "Preprocessing..."
     shared.state.job_count = len(files)
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index 5965c5a0..45397be9 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -161,6 +161,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
 
     shared.state.textinfo = "Initializing textual inversion training..."
     shared.state.job_count = steps
+    extns = [".jpg",".jpeg",".png"]
 
     filename = os.path.join(shared.cmd_opts.embeddings_dir, f'{embedding_name}.pt')
 
@@ -200,7 +201,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
     if ititial_step > steps:
         return embedding, filename
 
-    tr_img_len = len([os.path.join(data_root, file_path) for file_path in os.listdir(data_root)])
+    tr_img_len = len([os.path.join(data_root, file_path) for file_path in os.listdir(data_root) if os.path.splitext(file_path.casefold())[1] in extns])
     epoch_len = (tr_img_len * num_repeats) + tr_img_len
 
     pbar = tqdm.tqdm(enumerate(ds), total=steps-ititial_step)
-- 
cgit v1.2.3


From 907a88b2d0be320575c2129d8d6a1d4f3a68f9eb Mon Sep 17 00:00:00 2001
From: alg-wiki <alg.4chan@gmail.com>
Date: Tue, 11 Oct 2022 06:33:08 +0900
Subject: Added .webp .bmp

---
 modules/textual_inversion/dataset.py           | 2 +-
 modules/textual_inversion/preprocess.py        | 2 +-
 modules/textual_inversion/textual_inversion.py | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index d4baf066..0dc54fb7 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -22,7 +22,7 @@ class PersonalizedBase(Dataset):
         self.width = width
         self.height = height
         self.flip = transforms.RandomHorizontalFlip(p=flip_p)
-        self.extns = [".jpg",".jpeg",".png"]
+        self.extns = [".jpg",".jpeg",".png",".webp",".bmp"]
 
         self.dataset = []
 
diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py
index b6c78cf8..8290abe8 100644
--- a/modules/textual_inversion/preprocess.py
+++ b/modules/textual_inversion/preprocess.py
@@ -12,7 +12,7 @@ def preprocess(process_src, process_dst, process_width, process_height, process_
     height = process_height
     src = os.path.abspath(process_src)
     dst = os.path.abspath(process_dst)
-    extns = [".jpg",".jpeg",".png"]
+    extns = [".jpg",".jpeg",".png",".webp",".bmp"]
 
     assert src != dst, 'same directory specified as source and destination'
 
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index a03b299c..33c923d1 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -161,7 +161,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
 
     shared.state.textinfo = "Initializing textual inversion training..."
     shared.state.job_count = steps
-    extns = [".jpg",".jpeg",".png"]
+    extns = [".jpg",".jpeg",".png",".webp",".bmp"]
 
     filename = os.path.join(shared.cmd_opts.embeddings_dir, f'{embedding_name}.pt')
 
-- 
cgit v1.2.3


From b2368a3bce663f19a7209d9cb38617e635ca6e3c Mon Sep 17 00:00:00 2001
From: alg-wiki <alg.4chan@gmail.com>
Date: Tue, 11 Oct 2022 17:32:46 +0900
Subject: Switched to exception handling

---
 modules/textual_inversion/dataset.py           | 10 +++++-----
 modules/textual_inversion/preprocess.py        |  8 +++++---
 modules/textual_inversion/textual_inversion.py | 18 ++++++++----------
 3 files changed, 18 insertions(+), 18 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index 0dc54fb7..4d006366 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -22,7 +22,6 @@ class PersonalizedBase(Dataset):
         self.width = width
         self.height = height
         self.flip = transforms.RandomHorizontalFlip(p=flip_p)
-        self.extns = [".jpg",".jpeg",".png",".webp",".bmp"]
 
         self.dataset = []
 
@@ -33,12 +32,13 @@ class PersonalizedBase(Dataset):
 
         assert data_root, 'dataset directory not specified'
 
-        self.image_paths = [os.path.join(data_root, file_path) for file_path in os.listdir(data_root) if os.path.splitext(file_path.casefold())[1] in self.extns]
+        self.image_paths = [os.path.join(data_root, file_path) for file_path in os.listdir(data_root)]
         print("Preparing dataset...")
         for path in tqdm.tqdm(self.image_paths):
-            image = Image.open(path)
-            image = image.convert('RGB')
-            image = image.resize((self.width, self.height), PIL.Image.BICUBIC)
+            try:
+                image = Image.open(path).convert('RGB').resize((self.width, self.height), PIL.Image.BICUBIC)
+            except Exception:
+                continue
 
             filename = os.path.basename(path)
             filename_tokens = os.path.splitext(filename)[0]
diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py
index 8290abe8..1a672725 100644
--- a/modules/textual_inversion/preprocess.py
+++ b/modules/textual_inversion/preprocess.py
@@ -12,13 +12,12 @@ def preprocess(process_src, process_dst, process_width, process_height, process_
     height = process_height
     src = os.path.abspath(process_src)
     dst = os.path.abspath(process_dst)
-    extns = [".jpg",".jpeg",".png",".webp",".bmp"]
 
     assert src != dst, 'same directory specified as source and destination'
 
     os.makedirs(dst, exist_ok=True)
 
-    files = [i for i in os.listdir(src) if os.path.splitext(i.casefold())[1] in extns]
+    files = os.listdir(src)
 
     shared.state.textinfo = "Preprocessing..."
     shared.state.job_count = len(files)
@@ -47,7 +46,10 @@ def preprocess(process_src, process_dst, process_width, process_height, process_
     for index, imagefile in enumerate(tqdm.tqdm(files)):
         subindex = [0]
         filename = os.path.join(src, imagefile)
-        img = Image.open(filename).convert("RGB")
+        try:
+            img = Image.open(filename).convert("RGB")
+        except Exception:
+            continue
 
         if shared.state.interrupted:
             break
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index 33c923d1..91cde04b 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -161,7 +161,6 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
 
     shared.state.textinfo = "Initializing textual inversion training..."
     shared.state.job_count = steps
-    extns = [".jpg",".jpeg",".png",".webp",".bmp"]
 
     filename = os.path.join(shared.cmd_opts.embeddings_dir, f'{embedding_name}.pt')
 
@@ -201,10 +200,6 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
     if ititial_step > steps:
         return embedding, filename
 
-    tr_img_len = len([os.path.join(data_root, file_path) for file_path in os.listdir(data_root) if os.path.splitext(file_path.casefold())[1] in extns])
-
-    epoch_len = (tr_img_len * num_repeats) + tr_img_len
-
     pbar = tqdm.tqdm(enumerate(ds), total=steps-ititial_step)
     for i, (x, text) in pbar:
         embedding.step = i + ititial_step
@@ -228,10 +223,10 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
             loss.backward()
             optimizer.step()
 
-        epoch_num = embedding.step // epoch_len
-        epoch_step = embedding.step - (epoch_num * epoch_len) + 1
+        epoch_num = embedding.step // len(ds)
+        epoch_step = embedding.step - (epoch_num * len(ds)) + 1
 
-        pbar.set_description(f"[Epoch {epoch_num}: {epoch_step}/{epoch_len}]loss: {losses.mean():.7f}")
+        pbar.set_description(f"[Epoch {epoch_num}: {epoch_step}/{len(ds)}]loss: {losses.mean():.7f}")
 
         if embedding.step > 0 and embedding_dir is not None and embedding.step % save_embedding_every == 0:
             last_saved_file = os.path.join(embedding_dir, f'{embedding_name}-{embedding.step}.pt')
@@ -243,9 +238,12 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
             p = processing.StableDiffusionProcessingTxt2Img(
                 sd_model=shared.sd_model,
                 prompt=text,
-                steps=20,
-				height=training_height,
+                steps=28,
+				height=768,
 				width=training_width,
+                negative_prompt="lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts,signature, watermark, username, blurry, artist name",
+                cfg_scale=7.0,
+                sampler_index=0,
                 do_not_save_grid=True,
                 do_not_save_samples=True,
             )
-- 
cgit v1.2.3


From d4ea5f4d8631f778d11efcde397e4a5b8801d43b Mon Sep 17 00:00:00 2001
From: AUTOMATIC <16777216c@gmail.com>
Date: Tue, 11 Oct 2022 19:03:08 +0300
Subject: add an option to unload models during hypernetwork training to save
 VRAM

---
 modules/hypernetworks/hypernetwork.py          | 25 +++++++++++++++-------
 modules/hypernetworks/ui.py                    |  4 +++-
 modules/shared.py                              |  4 ++++
 modules/textual_inversion/dataset.py           | 29 ++++++++++++++++++--------
 modules/textual_inversion/textual_inversion.py |  2 +-
 5 files changed, 46 insertions(+), 18 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py
index b081f14e..4700e1ec 100644
--- a/modules/hypernetworks/hypernetwork.py
+++ b/modules/hypernetworks/hypernetwork.py
@@ -175,6 +175,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory,
     filename = os.path.join(shared.cmd_opts.hypernetwork_dir, f'{hypernetwork_name}.pt')
 
     log_directory = os.path.join(log_directory, datetime.datetime.now().strftime("%Y-%m-%d"), hypernetwork_name)
+    unload = shared.opts.unload_models_when_training
 
     if save_hypernetwork_every > 0:
         hypernetwork_dir = os.path.join(log_directory, "hypernetworks")
@@ -188,11 +189,13 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory,
     else:
         images_dir = None
 
-    cond_model = shared.sd_model.cond_stage_model
-
     shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..."
     with torch.autocast("cuda"):
-        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=512, height=512, repeats=1, placeholder_token=hypernetwork_name, model=shared.sd_model, device=devices.device, template_file=template_file)
+        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=512, height=512, repeats=1, placeholder_token=hypernetwork_name, model=shared.sd_model, device=devices.device, template_file=template_file, include_cond=True)
+
+    if unload:
+        shared.sd_model.cond_stage_model.to(devices.cpu)
+        shared.sd_model.first_stage_model.to(devices.cpu)
 
     hypernetwork = shared.loaded_hypernetwork
     weights = hypernetwork.weights()
@@ -211,7 +214,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory,
         return hypernetwork, filename
 
     pbar = tqdm.tqdm(enumerate(ds), total=steps - ititial_step)
-    for i, (x, text) in pbar:
+    for i, (x, text, cond) in pbar:
         hypernetwork.step = i + ititial_step
 
         if hypernetwork.step > steps:
@@ -221,11 +224,11 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory,
             break
 
         with torch.autocast("cuda"):
-            c = cond_model([text])
-
+            cond = cond.to(devices.device)
             x = x.to(devices.device)
-            loss = shared.sd_model(x.unsqueeze(0), c)[0]
+            loss = shared.sd_model(x.unsqueeze(0), cond)[0]
             del x
+            del cond
 
             losses[hypernetwork.step % losses.shape[0]] = loss.item()
 
@@ -244,6 +247,10 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory,
 
             preview_text = text if preview_image_prompt == "" else preview_image_prompt
 
+            optimizer.zero_grad()
+            shared.sd_model.cond_stage_model.to(devices.device)
+            shared.sd_model.first_stage_model.to(devices.device)
+
             p = processing.StableDiffusionProcessingTxt2Img(
                 sd_model=shared.sd_model,
                 prompt=preview_text,
@@ -255,6 +262,10 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory,
             processed = processing.process_images(p)
             image = processed.images[0]
 
+            if unload:
+                shared.sd_model.cond_stage_model.to(devices.cpu)
+                shared.sd_model.first_stage_model.to(devices.cpu)
+
             shared.state.current_image = image
             image.save(last_saved_image)
 
diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py
index 3541a388..c67facbb 100644
--- a/modules/hypernetworks/ui.py
+++ b/modules/hypernetworks/ui.py
@@ -5,7 +5,7 @@ import gradio as gr
 
 import modules.textual_inversion.textual_inversion
 import modules.textual_inversion.preprocess
-from modules import sd_hijack, shared
+from modules import sd_hijack, shared, devices
 from modules.hypernetworks import hypernetwork
 
 
@@ -41,5 +41,7 @@ Hypernetwork saved to {html.escape(filename)}
         raise
     finally:
         shared.loaded_hypernetwork = initial_hypernetwork
+        shared.sd_model.cond_stage_model.to(devices.device)
+        shared.sd_model.first_stage_model.to(devices.device)
         sd_hijack.apply_optimizations()
 
diff --git a/modules/shared.py b/modules/shared.py
index 20b45f23..c1092ff7 100644
--- a/modules/shared.py
+++ b/modules/shared.py
@@ -228,6 +228,10 @@ options_templates.update(options_section(('system', "System"), {
     "multiple_tqdm": OptionInfo(True, "Add a second progress bar to the console that shows progress for an entire job."),
 }))
 
+options_templates.update(options_section(('training', "Training"), {
+    "unload_models_when_training": OptionInfo(False, "Unload VAE and CLIP form VRAM when training"),
+}))
+
 options_templates.update(options_section(('sd', "Stable Diffusion"), {
     "sd_model_checkpoint": OptionInfo(None, "Stable Diffusion checkpoint", gr.Dropdown, lambda: {"choices": modules.sd_models.checkpoint_tiles()}, show_on_main_page=True),
     "sd_hypernetwork": OptionInfo("None", "Stable Diffusion finetune hypernetwork", gr.Dropdown, lambda: {"choices": ["None"] + [x for x in hypernetworks.keys()]}),
diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index 4d006366..f61f40d3 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -8,14 +8,14 @@ from torchvision import transforms
 
 import random
 import tqdm
-from modules import devices
+from modules import devices, shared
 import re
 
 re_tag = re.compile(r"[a-zA-Z][_\w\d()]+")
 
 
 class PersonalizedBase(Dataset):
-    def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", model=None, device=None, template_file=None):
+    def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", model=None, device=None, template_file=None, include_cond=False):
 
         self.placeholder_token = placeholder_token
 
@@ -32,6 +32,8 @@ class PersonalizedBase(Dataset):
 
         assert data_root, 'dataset directory not specified'
 
+        cond_model = shared.sd_model.cond_stage_model
+
         self.image_paths = [os.path.join(data_root, file_path) for file_path in os.listdir(data_root)]
         print("Preparing dataset...")
         for path in tqdm.tqdm(self.image_paths):
@@ -53,7 +55,13 @@ class PersonalizedBase(Dataset):
             init_latent = model.get_first_stage_encoding(model.encode_first_stage(torchdata.unsqueeze(dim=0))).squeeze()
             init_latent = init_latent.to(devices.cpu)
 
-            self.dataset.append((init_latent, filename_tokens))
+            if include_cond:
+                text = self.create_text(filename_tokens)
+                cond = cond_model([text]).to(devices.cpu)
+            else:
+                cond = None
+
+            self.dataset.append((init_latent, filename_tokens, cond))
 
         self.length = len(self.dataset) * repeats
 
@@ -64,6 +72,12 @@ class PersonalizedBase(Dataset):
     def shuffle(self):
         self.indexes = self.initial_indexes[torch.randperm(self.initial_indexes.shape[0])]
 
+    def create_text(self, filename_tokens):
+        text = random.choice(self.lines)
+        text = text.replace("[name]", self.placeholder_token)
+        text = text.replace("[filewords]", ' '.join(filename_tokens))
+        return text
+
     def __len__(self):
         return self.length
 
@@ -72,10 +86,7 @@ class PersonalizedBase(Dataset):
             self.shuffle()
 
         index = self.indexes[i % len(self.indexes)]
-        x, filename_tokens = self.dataset[index]
-
-        text = random.choice(self.lines)
-        text = text.replace("[name]", self.placeholder_token)
-        text = text.replace("[filewords]", ' '.join(filename_tokens))
+        x, filename_tokens, cond = self.dataset[index]
 
-        return x, text
+        text = self.create_text(filename_tokens)
+        return x, text, cond
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index bb05cdc6..35f4bd9e 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -201,7 +201,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
         return embedding, filename
 
     pbar = tqdm.tqdm(enumerate(ds), total=steps-ititial_step)
-    for i, (x, text) in pbar:
+    for i, (x, text, _) in pbar:
         embedding.step = i + ititial_step
 
         if embedding.step > steps:
-- 
cgit v1.2.3


From c3c8eef9fd5a0c8b26319e32ca4a19b56204e6df Mon Sep 17 00:00:00 2001
From: AUTOMATIC <16777216c@gmail.com>
Date: Wed, 12 Oct 2022 20:49:47 +0300
Subject: train: change filename processing to be more simple and configurable
 train: make it possible to make text files with prompts train: rework
 scheduler so that there's less repeating code in textual inversion and
 hypernets train: move epochs setting to options

---
 javascript/hints.js                            |  3 ++
 modules/hypernetworks/hypernetwork.py          | 40 +++++++++-------------
 modules/shared.py                              |  3 ++
 modules/textual_inversion/dataset.py           | 47 +++++++++++++++++++-------
 modules/textual_inversion/learn_schedule.py    | 37 +++++++++++++++++++-
 modules/textual_inversion/textual_inversion.py | 35 +++++++------------
 modules/ui.py                                  |  2 --
 7 files changed, 105 insertions(+), 62 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/javascript/hints.js b/javascript/hints.js
index b81c181b..d51ee14c 100644
--- a/javascript/hints.js
+++ b/javascript/hints.js
@@ -81,6 +81,9 @@ titles = {
 
     "Eta noise seed delta": "If this values is non-zero, it will be added to seed and used to initialize RNG for noises when using samplers with Eta. You can use this to produce even more variation of images, or you can use this to match images of other software if you know what you are doing.",
     "Do not add watermark to images": "If this option is enabled, watermark will not be added to created images. Warning: if you do not add watermark, you may be behaving in an unethical manner.",
+
+    "Filename word regex": "This regular expression will be used extract words from filename, and they will be joined using the option below into label text used for training. Leave empty to keep filename text as it is.",
+    "Filename join string": "This string will be used to hoin split words into a single line if the option above is enabled.",
 }
 
 
diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py
index 8314450a..b6c06d49 100644
--- a/modules/hypernetworks/hypernetwork.py
+++ b/modules/hypernetworks/hypernetwork.py
@@ -14,7 +14,7 @@ import torch
 from torch import einsum
 from einops import rearrange, repeat
 import modules.textual_inversion.dataset
-from modules.textual_inversion.learn_schedule import LearnSchedule
+from modules.textual_inversion.learn_schedule import LearnRateScheduler
 
 
 class HypernetworkModule(torch.nn.Module):
@@ -223,31 +223,23 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory,
     if ititial_step > steps:
         return hypernetwork, filename
 
-    schedules = iter(LearnSchedule(learn_rate, steps, ititial_step))
-    (learn_rate, end_step) = next(schedules)
-    print(f'Training at rate of {learn_rate} until step {end_step}')
-
-    optimizer = torch.optim.AdamW(weights, lr=learn_rate)
+    scheduler = LearnRateScheduler(learn_rate, steps, ititial_step)
+    optimizer = torch.optim.AdamW(weights, lr=scheduler.learn_rate)
 
     pbar = tqdm.tqdm(enumerate(ds), total=steps - ititial_step)
-    for i, (x, text, cond) in pbar:
+    for i, entry in pbar:
         hypernetwork.step = i + ititial_step
 
-        if hypernetwork.step > end_step:
-            try:
-                (learn_rate, end_step) = next(schedules)
-            except Exception:
-                break
-            tqdm.tqdm.write(f'Training at rate of {learn_rate} until step {end_step}')
-            for pg in optimizer.param_groups:
-                pg['lr'] = learn_rate
+        scheduler.apply(optimizer, hypernetwork.step)
+        if scheduler.finished:
+            break
 
         if shared.state.interrupted:
             break
 
         with torch.autocast("cuda"):
-            cond = cond.to(devices.device)
-            x = x.to(devices.device)
+            cond = entry.cond.to(devices.device)
+            x = entry.latent.to(devices.device)
             loss = shared.sd_model(x.unsqueeze(0), cond)[0]
             del x
             del cond
@@ -267,7 +259,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory,
         if hypernetwork.step > 0 and images_dir is not None and hypernetwork.step % create_image_every == 0:
             last_saved_image = os.path.join(images_dir, f'{hypernetwork_name}-{hypernetwork.step}.png')
 
-            preview_text = text if preview_image_prompt == "" else preview_image_prompt
+            preview_text = entry.cond_text if preview_image_prompt == "" else preview_image_prompt
 
             optimizer.zero_grad()
             shared.sd_model.cond_stage_model.to(devices.device)
@@ -282,16 +274,16 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory,
             )
 
             processed = processing.process_images(p)
-            image = processed.images[0]
+            image = processed.images[0] if len(processed.images)>0 else None
 
             if unload:
                 shared.sd_model.cond_stage_model.to(devices.cpu)
                 shared.sd_model.first_stage_model.to(devices.cpu)
 
-            shared.state.current_image = image
-            image.save(last_saved_image)
-
-            last_saved_image += f", prompt: {preview_text}"
+            if image is not None:
+                shared.state.current_image = image
+                image.save(last_saved_image)
+                last_saved_image += f", prompt: {preview_text}"
 
         shared.state.job_no = hypernetwork.step
 
@@ -299,7 +291,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory,
 <p>
 Loss: {losses.mean():.7f}<br/>
 Step: {hypernetwork.step}<br/>
-Last prompt: {html.escape(text)}<br/>
+Last prompt: {html.escape(entry.cond_text)}<br/>
 Last saved embedding: {html.escape(last_saved_file)}<br/>
 Last saved image: {html.escape(last_saved_image)}<br/>
 </p>
diff --git a/modules/shared.py b/modules/shared.py
index 42e99741..e64e69fc 100644
--- a/modules/shared.py
+++ b/modules/shared.py
@@ -231,6 +231,9 @@ options_templates.update(options_section(('system', "System"), {
 
 options_templates.update(options_section(('training', "Training"), {
     "unload_models_when_training": OptionInfo(False, "Unload VAE and CLIP from VRAM when training"),
+    "dataset_filename_word_regex": OptionInfo("", "Filename word regex"),
+    "dataset_filename_join_string": OptionInfo(" ", "Filename join string"),
+    "training_image_repeats_per_epoch": OptionInfo(100, "Number of repeats for a single input image per epoch; used only for displaying epoch number", gr.Number, {"precision": 0}),
 }))
 
 options_templates.update(options_section(('sd', "Stable Diffusion"), {
diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index f61f40d3..67e90afe 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -11,11 +11,21 @@ import tqdm
 from modules import devices, shared
 import re
 
-re_tag = re.compile(r"[a-zA-Z][_\w\d()]+")
+re_numbers_at_start = re.compile(r"^[-\d]+\s*")
+
+
+class DatasetEntry:
+    def __init__(self, filename=None, latent=None, filename_text=None):
+        self.filename = filename
+        self.latent = latent
+        self.filename_text = filename_text
+        self.cond = None
+        self.cond_text = None
 
 
 class PersonalizedBase(Dataset):
     def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", model=None, device=None, template_file=None, include_cond=False):
+        re_word = re.compile(shared.opts.dataset_filename_word_regex) if len(shared.opts.dataset_filename_word_regex)>0 else None
 
         self.placeholder_token = placeholder_token
 
@@ -42,9 +52,18 @@ class PersonalizedBase(Dataset):
             except Exception:
                 continue
 
+            text_filename = os.path.splitext(path)[0] + ".txt"
             filename = os.path.basename(path)
-            filename_tokens = os.path.splitext(filename)[0]
-            filename_tokens = re_tag.findall(filename_tokens)
+
+            if os.path.exists(text_filename):
+                with open(text_filename, "r", encoding="utf8") as file:
+                    filename_text = file.read()
+            else:
+                filename_text = os.path.splitext(filename)[0]
+                filename_text = re.sub(re_numbers_at_start, '', filename_text)
+                if re_word:
+                    tokens = re_word.findall(filename_text)
+                    filename_text = (shared.opts.dataset_filename_join_string or "").join(tokens)
 
             npimage = np.array(image).astype(np.uint8)
             npimage = (npimage / 127.5 - 1.0).astype(np.float32)
@@ -55,13 +74,13 @@ class PersonalizedBase(Dataset):
             init_latent = model.get_first_stage_encoding(model.encode_first_stage(torchdata.unsqueeze(dim=0))).squeeze()
             init_latent = init_latent.to(devices.cpu)
 
+            entry = DatasetEntry(filename=path, filename_text=filename_text, latent=init_latent)
+
             if include_cond:
-                text = self.create_text(filename_tokens)
-                cond = cond_model([text]).to(devices.cpu)
-            else:
-                cond = None
+                entry.cond_text = self.create_text(filename_text)
+                entry.cond = cond_model([entry.cond_text]).to(devices.cpu)
 
-            self.dataset.append((init_latent, filename_tokens, cond))
+            self.dataset.append(entry)
 
         self.length = len(self.dataset) * repeats
 
@@ -72,10 +91,10 @@ class PersonalizedBase(Dataset):
     def shuffle(self):
         self.indexes = self.initial_indexes[torch.randperm(self.initial_indexes.shape[0])]
 
-    def create_text(self, filename_tokens):
+    def create_text(self, filename_text):
         text = random.choice(self.lines)
         text = text.replace("[name]", self.placeholder_token)
-        text = text.replace("[filewords]", ' '.join(filename_tokens))
+        text = text.replace("[filewords]", filename_text)
         return text
 
     def __len__(self):
@@ -86,7 +105,9 @@ class PersonalizedBase(Dataset):
             self.shuffle()
 
         index = self.indexes[i % len(self.indexes)]
-        x, filename_tokens, cond = self.dataset[index]
+        entry = self.dataset[index]
+
+        if entry.cond is None:
+            entry.cond_text = self.create_text(entry.filename_text)
 
-        text = self.create_text(filename_tokens)
-        return x, text, cond
+        return entry
diff --git a/modules/textual_inversion/learn_schedule.py b/modules/textual_inversion/learn_schedule.py
index db720271..2062726a 100644
--- a/modules/textual_inversion/learn_schedule.py
+++ b/modules/textual_inversion/learn_schedule.py
@@ -1,6 +1,12 @@
+import tqdm
 
-class LearnSchedule:
+
+class LearnScheduleIterator:
     def __init__(self, learn_rate, max_steps, cur_step=0):
+        """
+        specify learn_rate as "0.001:100, 0.00001:1000, 1e-5:10000" to have lr of 0.001 until step 100, 0.00001 until 1000, 1e-5:10000 until 10000
+        """
+
         pairs = learn_rate.split(',')
         self.rates = []
         self.it = 0
@@ -32,3 +38,32 @@ class LearnSchedule:
             return self.rates[self.it - 1]
         else:
             raise StopIteration
+
+
+class LearnRateScheduler:
+    def __init__(self, learn_rate, max_steps, cur_step=0, verbose=True):
+        self.schedules = LearnScheduleIterator(learn_rate, max_steps, cur_step)
+        (self.learn_rate,  self.end_step) = next(self.schedules)
+        self.verbose = verbose
+
+        if self.verbose:
+            print(f'Training at rate of {self.learn_rate} until step {self.end_step}')
+
+        self.finished = False
+
+    def apply(self, optimizer, step_number):
+        if step_number <= self.end_step:
+            return
+
+        try:
+            (self.learn_rate, self.end_step) = next(self.schedules)
+        except Exception:
+            self.finished = True
+            return
+
+        if self.verbose:
+            tqdm.tqdm.write(f'Training at rate of {self.learn_rate} until step {self.end_step}')
+
+        for pg in optimizer.param_groups:
+            pg['lr'] = self.learn_rate
+
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index c5153e4a..fa0e33a2 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -11,7 +11,7 @@ from PIL import Image, PngImagePlugin
 
 from modules import shared, devices, sd_hijack, processing, sd_models
 import modules.textual_inversion.dataset
-from modules.textual_inversion.learn_schedule import LearnSchedule
+from modules.textual_inversion.learn_schedule import LearnRateScheduler
 
 from modules.textual_inversion.image_embedding import (embedding_to_b64, embedding_from_b64,
                                                        insert_image_data_embed, extract_image_data_embed,
@@ -172,8 +172,7 @@ def create_embedding(name, num_vectors_per_token, init_text='*'):
     return fn
 
 
-
-def train_embedding(embedding_name, learn_rate, data_root, log_directory, training_width, training_height, steps, num_repeats, create_image_every, save_embedding_every, template_file, save_image_with_stored_embedding, preview_image_prompt):
+def train_embedding(embedding_name, learn_rate, data_root, log_directory, training_width, training_height, steps, create_image_every, save_embedding_every, template_file, save_image_with_stored_embedding, preview_image_prompt):
     assert embedding_name, 'embedding not selected'
 
     shared.state.textinfo = "Initializing textual inversion training..."
@@ -205,7 +204,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
 
     shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..."
     with torch.autocast("cuda"):
-        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=num_repeats, placeholder_token=embedding_name, model=shared.sd_model, device=devices.device, template_file=template_file)
+        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=embedding_name, model=shared.sd_model, device=devices.device, template_file=template_file)
 
     hijack = sd_hijack.model_hijack
 
@@ -221,32 +220,24 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
     if ititial_step > steps:
         return embedding, filename
 
-    schedules = iter(LearnSchedule(learn_rate, steps, ititial_step))
-    (learn_rate, end_step) = next(schedules)
-    print(f'Training at rate of {learn_rate} until step {end_step}')
-
-    optimizer = torch.optim.AdamW([embedding.vec], lr=learn_rate)
+    scheduler = LearnRateScheduler(learn_rate, steps, ititial_step)
+    optimizer = torch.optim.AdamW([embedding.vec], lr=scheduler.learn_rate)
 
     pbar = tqdm.tqdm(enumerate(ds), total=steps-ititial_step)
-    for i, (x, text, _) in pbar:
+    for i, entry in pbar:
         embedding.step = i + ititial_step
 
-        if embedding.step > end_step:
-            try:
-                (learn_rate, end_step) = next(schedules)
-            except:
-                break
-            tqdm.tqdm.write(f'Training at rate of {learn_rate} until step {end_step}')
-            for pg in optimizer.param_groups:
-                pg['lr'] = learn_rate
+        scheduler.apply(optimizer, embedding.step)
+        if scheduler.finished:
+            break
 
         if shared.state.interrupted:
             break
 
         with torch.autocast("cuda"):
-            c = cond_model([text])
+            c = cond_model([entry.cond_text])
 
-            x = x.to(devices.device)
+            x = entry.latent.to(devices.device)
             loss = shared.sd_model(x.unsqueeze(0), c)[0]
             del x
 
@@ -268,7 +259,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
         if embedding.step > 0 and images_dir is not None and embedding.step % create_image_every == 0:
             last_saved_image = os.path.join(images_dir, f'{embedding_name}-{embedding.step}.png')
 
-            preview_text = text if preview_image_prompt == "" else preview_image_prompt
+            preview_text = entry.cond_text if preview_image_prompt == "" else preview_image_prompt
 
             p = processing.StableDiffusionProcessingTxt2Img(
                 sd_model=shared.sd_model,
@@ -314,7 +305,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
 <p>
 Loss: {losses.mean():.7f}<br/>
 Step: {embedding.step}<br/>
-Last prompt: {html.escape(text)}<br/>
+Last prompt: {html.escape(entry.cond_text)}<br/>
 Last saved embedding: {html.escape(last_saved_file)}<br/>
 Last saved image: {html.escape(last_saved_image)}<br/>
 </p>
diff --git a/modules/ui.py b/modules/ui.py
index 2b332267..c42535c8 100644
--- a/modules/ui.py
+++ b/modules/ui.py
@@ -1098,7 +1098,6 @@ def create_ui(wrap_gradio_gpu_call):
                     training_width = gr.Slider(minimum=64, maximum=2048, step=64, label="Width", value=512)
                     training_height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height", value=512)
                     steps = gr.Number(label='Max steps', value=100000, precision=0)
-                    num_repeats = gr.Number(label='Number of repeats for a single input image per epoch', value=100, precision=0)
                     create_image_every = gr.Number(label='Save an image to log directory every N steps, 0 to disable', value=500, precision=0)
                     save_embedding_every = gr.Number(label='Save a copy of embedding to log directory every N steps, 0 to disable', value=500, precision=0)
                     save_image_with_stored_embedding = gr.Checkbox(label='Save images with embedding in PNG chunks', value=True)
@@ -1176,7 +1175,6 @@ def create_ui(wrap_gradio_gpu_call):
                 training_width,
                 training_height,
                 steps,
-                num_repeats,
                 create_image_every,
                 save_embedding_every,
                 template_file,
-- 
cgit v1.2.3


From 4d19f3b7d461fe0f63e7ccff936909b0ce0c6126 Mon Sep 17 00:00:00 2001
From: Melan <alexleander91@gmail.com>
Date: Fri, 14 Oct 2022 22:45:26 +0200
Subject: Raise an assertion error if no training images have been found.

---
 modules/textual_inversion/dataset.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index 67e90afe..12e2f43b 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -81,7 +81,8 @@ class PersonalizedBase(Dataset):
                 entry.cond = cond_model([entry.cond_text]).to(devices.cpu)
 
             self.dataset.append(entry)
-
+        
+        assert len(self.dataset) > 1, "No images have been found in the dataset."
         self.length = len(self.dataset) * repeats
 
         self.initial_indexes = np.arange(self.length) % len(self.dataset)
-- 
cgit v1.2.3


From c7a86f7fe9c0b8967a87e8d709f507d2f44400d8 Mon Sep 17 00:00:00 2001
From: AUTOMATIC <16777216c@gmail.com>
Date: Sat, 15 Oct 2022 09:24:59 +0300
Subject: add option to use batch size for training

---
 modules/hypernetworks/hypernetwork.py          | 33 +++++++++++++++++++-------
 modules/textual_inversion/dataset.py           | 31 ++++++++++++++----------
 modules/textual_inversion/textual_inversion.py | 17 +++++++------
 modules/ui.py                                  |  3 +++
 4 files changed, 54 insertions(+), 30 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py
index 59c7ac6e..a2b3bc0a 100644
--- a/modules/hypernetworks/hypernetwork.py
+++ b/modules/hypernetworks/hypernetwork.py
@@ -182,7 +182,21 @@ def attention_CrossAttention_forward(self, x, context=None, mask=None):
     return self.to_out(out)
 
 
-def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory, steps, create_image_every, save_hypernetwork_every, template_file, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
+def stack_conds(conds):
+    if len(conds) == 1:
+        return torch.stack(conds)
+
+    # same as in reconstruct_multicond_batch
+    token_count = max([x.shape[0] for x in conds])
+    for i in range(len(conds)):
+        if conds[i].shape[0] != token_count:
+            last_vector = conds[i][-1:]
+            last_vector_repeated = last_vector.repeat([token_count - conds[i].shape[0], 1])
+            conds[i] = torch.vstack([conds[i], last_vector_repeated])
+
+    return torch.stack(conds)
+
+def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log_directory, steps, create_image_every, save_hypernetwork_every, template_file, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
     assert hypernetwork_name, 'hypernetwork not selected'
 
     path = shared.hypernetworks.get(hypernetwork_name, None)
@@ -211,7 +225,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory,
 
     shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..."
     with torch.autocast("cuda"):
-        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=512, height=512, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=hypernetwork_name, model=shared.sd_model, device=devices.device, template_file=template_file, include_cond=True)
+        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=512, height=512, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=hypernetwork_name, model=shared.sd_model, device=devices.device, template_file=template_file, include_cond=True, batch_size=batch_size)
 
     if unload:
         shared.sd_model.cond_stage_model.to(devices.cpu)
@@ -235,7 +249,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory,
     optimizer = torch.optim.AdamW(weights, lr=scheduler.learn_rate)
 
     pbar = tqdm.tqdm(enumerate(ds), total=steps - ititial_step)
-    for i, entry in pbar:
+    for i, entries in pbar:
         hypernetwork.step = i + ititial_step
 
         scheduler.apply(optimizer, hypernetwork.step)
@@ -246,11 +260,12 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory,
             break
 
         with torch.autocast("cuda"):
-            cond = entry.cond.to(devices.device)
-            x = entry.latent.to(devices.device)
-            loss = shared.sd_model(x.unsqueeze(0), cond)[0]
+            c = stack_conds([entry.cond for entry in entries]).to(devices.device)
+#            c = torch.vstack([entry.cond for entry in entries]).to(devices.device)
+            x = torch.stack([entry.latent for entry in entries]).to(devices.device)
+            loss = shared.sd_model(x, c)[0]
             del x
-            del cond
+            del c
 
             losses[hypernetwork.step % losses.shape[0]] = loss.item()
 
@@ -292,7 +307,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory,
                 p.width = preview_width
                 p.height = preview_height
             else:
-                p.prompt = entry.cond_text
+                p.prompt = entries[0].cond_text
                 p.steps = 20
 
             preview_text = p.prompt
@@ -315,7 +330,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory,
 <p>
 Loss: {losses.mean():.7f}<br/>
 Step: {hypernetwork.step}<br/>
-Last prompt: {html.escape(entry.cond_text)}<br/>
+Last prompt: {html.escape(entries[0].cond_text)}<br/>
 Last saved embedding: {html.escape(last_saved_file)}<br/>
 Last saved image: {html.escape(last_saved_image)}<br/>
 </p>
diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index 67e90afe..bd99c0cb 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -24,11 +24,12 @@ class DatasetEntry:
 
 
 class PersonalizedBase(Dataset):
-    def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", model=None, device=None, template_file=None, include_cond=False):
-        re_word = re.compile(shared.opts.dataset_filename_word_regex) if len(shared.opts.dataset_filename_word_regex)>0 else None
+    def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", model=None, device=None, template_file=None, include_cond=False, batch_size=1):
+        re_word = re.compile(shared.opts.dataset_filename_word_regex) if len(shared.opts.dataset_filename_word_regex) > 0 else None
 
         self.placeholder_token = placeholder_token
 
+        self.batch_size = batch_size
         self.width = width
         self.height = height
         self.flip = transforms.RandomHorizontalFlip(p=flip_p)
@@ -78,13 +79,13 @@ class PersonalizedBase(Dataset):
 
             if include_cond:
                 entry.cond_text = self.create_text(filename_text)
-                entry.cond = cond_model([entry.cond_text]).to(devices.cpu)
+                entry.cond = cond_model([entry.cond_text]).to(devices.cpu).squeeze(0)
 
             self.dataset.append(entry)
 
-        self.length = len(self.dataset) * repeats
+        self.length = len(self.dataset) * repeats // batch_size
 
-        self.initial_indexes = np.arange(self.length) % len(self.dataset)
+        self.initial_indexes = np.arange(len(self.dataset))
         self.indexes = None
         self.shuffle()
 
@@ -101,13 +102,19 @@ class PersonalizedBase(Dataset):
         return self.length
 
     def __getitem__(self, i):
-        if i % len(self.dataset) == 0:
-            self.shuffle()
+        res = []
 
-        index = self.indexes[i % len(self.indexes)]
-        entry = self.dataset[index]
+        for j in range(self.batch_size):
+            position = i * self.batch_size + j
+            if position % len(self.indexes) == 0:
+                self.shuffle()
 
-        if entry.cond is None:
-            entry.cond_text = self.create_text(entry.filename_text)
+            index = self.indexes[position % len(self.indexes)]
+            entry = self.dataset[index]
 
-        return entry
+            if entry.cond is None:
+                entry.cond_text = self.create_text(entry.filename_text)
+
+            res.append(entry)
+
+        return res
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index da0d77a0..e754747e 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -199,7 +199,7 @@ def write_loss(log_directory, filename, step, epoch_len, values):
         })
 
 
-def train_embedding(embedding_name, learn_rate, data_root, log_directory, training_width, training_height, steps, create_image_every, save_embedding_every, template_file, save_image_with_stored_embedding, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
+def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_directory, training_width, training_height, steps, create_image_every, save_embedding_every, template_file, save_image_with_stored_embedding, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
     assert embedding_name, 'embedding not selected'
 
     shared.state.textinfo = "Initializing textual inversion training..."
@@ -231,7 +231,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
 
     shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..."
     with torch.autocast("cuda"):
-        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=embedding_name, model=shared.sd_model, device=devices.device, template_file=template_file)
+        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=embedding_name, model=shared.sd_model, device=devices.device, template_file=template_file, batch_size=batch_size)
 
     hijack = sd_hijack.model_hijack
 
@@ -251,7 +251,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
     optimizer = torch.optim.AdamW([embedding.vec], lr=scheduler.learn_rate)
 
     pbar = tqdm.tqdm(enumerate(ds), total=steps-ititial_step)
-    for i, entry in pbar:
+    for i, entries in pbar:
         embedding.step = i + ititial_step
 
         scheduler.apply(optimizer, embedding.step)
@@ -262,10 +262,9 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
             break
 
         with torch.autocast("cuda"):
-            c = cond_model([entry.cond_text])
-
-            x = entry.latent.to(devices.device)
-            loss = shared.sd_model(x.unsqueeze(0), c)[0]
+            c = cond_model([entry.cond_text for entry in entries])
+            x = torch.stack([entry.latent for entry in entries]).to(devices.device)
+            loss = shared.sd_model(x, c)[0]
             del x
 
             losses[embedding.step % losses.shape[0]] = loss.item()
@@ -307,7 +306,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
                 p.width = preview_width
                 p.height = preview_height
             else:
-                p.prompt = entry.cond_text
+                p.prompt = entries[0].cond_text
                 p.steps = 20
                 p.width = training_width
                 p.height = training_height
@@ -348,7 +347,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini
 <p>
 Loss: {losses.mean():.7f}<br/>
 Step: {embedding.step}<br/>
-Last prompt: {html.escape(entry.cond_text)}<br/>
+Last prompt: {html.escape(entries[0].cond_text)}<br/>
 Last saved embedding: {html.escape(last_saved_file)}<br/>
 Last saved image: {html.escape(last_saved_image)}<br/>
 </p>
diff --git a/modules/ui.py b/modules/ui.py
index 1bc919c7..45550ea8 100644
--- a/modules/ui.py
+++ b/modules/ui.py
@@ -1166,6 +1166,7 @@ def create_ui(wrap_gradio_gpu_call):
                     train_embedding_name = gr.Dropdown(label='Embedding', choices=sorted(sd_hijack.model_hijack.embedding_db.word_embeddings.keys()))
                     train_hypernetwork_name = gr.Dropdown(label='Hypernetwork', choices=[x for x in shared.hypernetworks.keys()])
                     learn_rate = gr.Textbox(label='Learning rate', placeholder="Learning rate", value="0.005")
+                    batch_size = gr.Number(label='Batch size', value=1, precision=0)
                     dataset_directory = gr.Textbox(label='Dataset directory', placeholder="Path to directory with input images")
                     log_directory = gr.Textbox(label='Log directory', placeholder="Path to directory where to write outputs", value="textual_inversion")
                     template_file = gr.Textbox(label='Prompt template file', value=os.path.join(script_path, "textual_inversion_templates", "style_filewords.txt"))
@@ -1244,6 +1245,7 @@ def create_ui(wrap_gradio_gpu_call):
             inputs=[
                 train_embedding_name,
                 learn_rate,
+                batch_size,
                 dataset_directory,
                 log_directory,
                 training_width,
@@ -1268,6 +1270,7 @@ def create_ui(wrap_gradio_gpu_call):
             inputs=[
                 train_hypernetwork_name,
                 learn_rate,
+                batch_size,
                 dataset_directory,
                 log_directory,
                 steps,
-- 
cgit v1.2.3


From b69c37d25e4ffc56e8f8c247fa2c38b4648cefb7 Mon Sep 17 00:00:00 2001
From: guaneec <guaneec@users.noreply.github.com>
Date: Thu, 20 Oct 2022 22:21:12 +0800
Subject: Allow datasets with only 1 image in TI

---
 modules/textual_inversion/dataset.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index 23bb4b6a..5b1c5002 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -83,7 +83,7 @@ class PersonalizedBase(Dataset):
 
             self.dataset.append(entry)
 
-        assert len(self.dataset) > 1, "No images have been found in the dataset."
+        assert len(self.dataset) > 0, "No images have been found in the dataset."
         self.length = len(self.dataset) * repeats // batch_size
 
         self.initial_indexes = np.arange(len(self.dataset))
@@ -91,7 +91,7 @@ class PersonalizedBase(Dataset):
         self.shuffle()
 
     def shuffle(self):
-        self.indexes = self.initial_indexes[torch.randperm(self.initial_indexes.shape[0])]
+        self.indexes = self.initial_indexes[torch.randperm(self.initial_indexes.shape[0]).numpy()]
 
     def create_text(self, filename_text):
         text = random.choice(self.lines)
-- 
cgit v1.2.3


From a0a7024c679056dd66beb1832e52041b10143130 Mon Sep 17 00:00:00 2001
From: FlameLaw <116745066+FlameLaw@users.noreply.github.com>
Date: Fri, 28 Oct 2022 02:13:48 +0900
Subject: Fix random dataset shuffle on TI

---
 modules/textual_inversion/dataset.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index 5b1c5002..8bb00d27 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -86,12 +86,12 @@ class PersonalizedBase(Dataset):
         assert len(self.dataset) > 0, "No images have been found in the dataset."
         self.length = len(self.dataset) * repeats // batch_size
 
-        self.initial_indexes = np.arange(len(self.dataset))
+        self.dataset_length = len(self.dataset)
         self.indexes = None
         self.shuffle()
 
     def shuffle(self):
-        self.indexes = self.initial_indexes[torch.randperm(self.initial_indexes.shape[0]).numpy()]
+        self.indexes = np.random.permutation(self.dataset_length)
 
     def create_text(self, filename_text):
         text = random.choice(self.lines)
-- 
cgit v1.2.3


From a27d19de2eff633b6a39f9f4a5c0f2d6abb81bb5 Mon Sep 17 00:00:00 2001
From: Muhammad Rizqi Nur <rizqinur2010@gmail.com>
Date: Sat, 29 Oct 2022 19:44:05 +0700
Subject: Additional assert on dataset

---
 modules/textual_inversion/dataset.py | 2 ++
 1 file changed, 2 insertions(+)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index 8bb00d27..ad726577 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -42,6 +42,8 @@ class PersonalizedBase(Dataset):
         self.lines = lines
 
         assert data_root, 'dataset directory not specified'
+        assert os.path.isdir(data_root), "Dataset directory doesn't exist"
+        assert os.listdir(data_root), "Dataset directory is empty"
 
         cond_model = shared.sd_model.cond_stage_model
 
-- 
cgit v1.2.3


From 467cae167a3066ffa2b2a5e6f16dd42642219aba Mon Sep 17 00:00:00 2001
From: TinkTheBoush <TinkTheBoush@github.com>
Date: Tue, 1 Nov 2022 23:29:12 +0900
Subject: append_tag_shuffle

---
 modules/hypernetworks/hypernetwork.py          |  4 ++--
 modules/textual_inversion/dataset.py           | 10 ++++++++--
 modules/textual_inversion/textual_inversion.py |  4 ++--
 modules/ui.py                                  |  3 +++
 4 files changed, 15 insertions(+), 6 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py
index a11e01d6..7630fb81 100644
--- a/modules/hypernetworks/hypernetwork.py
+++ b/modules/hypernetworks/hypernetwork.py
@@ -331,7 +331,7 @@ def report_statistics(loss_info:dict):
 
 
 
-def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log_directory, training_width, training_height, steps, create_image_every, save_hypernetwork_every, template_file, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
+def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log_directory, training_width, training_height, steps, create_image_every, save_hypernetwork_every, template_file, preview_from_txt2img, shuffle_tags, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
     # images allows training previews to have infotext. Importing it at the top causes a circular import problem.
     from modules import images
 
@@ -376,7 +376,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
     # dataset loading may take a while, so input validations and early returns should be done before this
     shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..."
     with torch.autocast("cuda"):
-        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=hypernetwork_name, model=shared.sd_model, device=devices.device, template_file=template_file, include_cond=True, batch_size=batch_size)
+        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=hypernetwork_name, shuffle_tags=shuffle_tags, model=shared.sd_model, device=devices.device, template_file=template_file, include_cond=True, batch_size=batch_size)
 
     if unload:
         shared.sd_model.cond_stage_model.to(devices.cpu)
diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index ad726577..e9d97cc1 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -24,7 +24,7 @@ class DatasetEntry:
 
 
 class PersonalizedBase(Dataset):
-    def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", model=None, device=None, template_file=None, include_cond=False, batch_size=1):
+    def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", shuffle_tags=True, model=None, device=None, template_file=None, include_cond=False, batch_size=1):
         re_word = re.compile(shared.opts.dataset_filename_word_regex) if len(shared.opts.dataset_filename_word_regex) > 0 else None
 
         self.placeholder_token = placeholder_token
@@ -33,6 +33,7 @@ class PersonalizedBase(Dataset):
         self.width = width
         self.height = height
         self.flip = transforms.RandomHorizontalFlip(p=flip_p)
+        self.shuffle_tags = shuffle_tags
 
         self.dataset = []
 
@@ -98,7 +99,12 @@ class PersonalizedBase(Dataset):
     def create_text(self, filename_text):
         text = random.choice(self.lines)
         text = text.replace("[name]", self.placeholder_token)
-        text = text.replace("[filewords]", filename_text)
+        if self.tag_shuffle:
+            tags = filename_text.split(',')
+            random.shuffle(tags)
+            text = text.replace("[filewords]", ','.join(tags))
+        else:
+            text = text.replace("[filewords]", filename_text)
         return text
 
     def __len__(self):
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index e0babb46..64700e23 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -224,7 +224,7 @@ def validate_train_inputs(model_name, learn_rate, batch_size, data_root, templat
     if save_model_every or create_image_every:
         assert log_directory, "Log directory is empty"
 
-def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_directory, training_width, training_height, steps, create_image_every, save_embedding_every, template_file, save_image_with_stored_embedding, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
+def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_directory, training_width, training_height, steps, create_image_every, save_embedding_every, template_file, save_image_with_stored_embedding, preview_from_txt2img, shuffle_tags, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
     save_embedding_every = save_embedding_every or 0
     create_image_every = create_image_every or 0
     validate_train_inputs(embedding_name, learn_rate, batch_size, data_root, template_file, steps, save_embedding_every, create_image_every, log_directory, name="embedding")
@@ -271,7 +271,7 @@ def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_direc
     # dataset loading may take a while, so input validations and early returns should be done before this
     shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..."
     with torch.autocast("cuda"):
-        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=embedding_name, model=shared.sd_model, device=devices.device, template_file=template_file, batch_size=batch_size)
+        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=embedding_name, shuffle_tags=shuffle_tags, model=shared.sd_model, device=devices.device, template_file=template_file, batch_size=batch_size)
 
     embedding.vec.requires_grad = True
     optimizer = torch.optim.AdamW([embedding.vec], lr=scheduler.learn_rate)
diff --git a/modules/ui.py b/modules/ui.py
index 2c15abb7..ad383979 100644
--- a/modules/ui.py
+++ b/modules/ui.py
@@ -1267,6 +1267,7 @@ def create_ui(wrap_gradio_gpu_call):
                     save_embedding_every = gr.Number(label='Save a copy of embedding to log directory every N steps, 0 to disable', value=500, precision=0)
                     save_image_with_stored_embedding = gr.Checkbox(label='Save images with embedding in PNG chunks', value=True)
                     preview_from_txt2img = gr.Checkbox(label='Read parameters (prompt, etc...) from txt2img tab when making previews', value=False)
+                    shuffle_tags = gr.Checkbox(label='Shuffleing tags by "," when create texts', value=True)
 
                     with gr.Row():
                         interrupt_training = gr.Button(value="Interrupt")
@@ -1361,6 +1362,7 @@ def create_ui(wrap_gradio_gpu_call):
                 template_file,
                 save_image_with_stored_embedding,
                 preview_from_txt2img,
+                shuffle_tags,
                 *txt2img_preview_params,
             ],
             outputs=[
@@ -1385,6 +1387,7 @@ def create_ui(wrap_gradio_gpu_call):
                 save_embedding_every,
                 template_file,
                 preview_from_txt2img,
+                shuffle_tags,
                 *txt2img_preview_params,
             ],
             outputs=[
-- 
cgit v1.2.3


From 821e2b883dbb42a187bc37379175cd55b7cd7e81 Mon Sep 17 00:00:00 2001
From: TinkTheBoush <TinkTheBoush@github.com>
Date: Fri, 4 Nov 2022 19:39:03 +0900
Subject: change option position to Training setting

---
 modules/hypernetworks/hypernetwork.py          | 4 ++--
 modules/shared.py                              | 1 +
 modules/textual_inversion/dataset.py           | 5 ++---
 modules/textual_inversion/textual_inversion.py | 4 ++--
 4 files changed, 7 insertions(+), 7 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py
index 7630fb81..a11e01d6 100644
--- a/modules/hypernetworks/hypernetwork.py
+++ b/modules/hypernetworks/hypernetwork.py
@@ -331,7 +331,7 @@ def report_statistics(loss_info:dict):
 
 
 
-def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log_directory, training_width, training_height, steps, create_image_every, save_hypernetwork_every, template_file, preview_from_txt2img, shuffle_tags, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
+def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log_directory, training_width, training_height, steps, create_image_every, save_hypernetwork_every, template_file, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
     # images allows training previews to have infotext. Importing it at the top causes a circular import problem.
     from modules import images
 
@@ -376,7 +376,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
     # dataset loading may take a while, so input validations and early returns should be done before this
     shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..."
     with torch.autocast("cuda"):
-        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=hypernetwork_name, shuffle_tags=shuffle_tags, model=shared.sd_model, device=devices.device, template_file=template_file, include_cond=True, batch_size=batch_size)
+        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=hypernetwork_name, model=shared.sd_model, device=devices.device, template_file=template_file, include_cond=True, batch_size=batch_size)
 
     if unload:
         shared.sd_model.cond_stage_model.to(devices.cpu)
diff --git a/modules/shared.py b/modules/shared.py
index 1ccb269a..e1d9bdf1 100644
--- a/modules/shared.py
+++ b/modules/shared.py
@@ -290,6 +290,7 @@ options_templates.update(options_section(('system', "System"), {
 
 options_templates.update(options_section(('training', "Training"), {
     "unload_models_when_training": OptionInfo(False, "Move VAE and CLIP to RAM when training if possible. Saves VRAM."),
+    "shuffle_tags": OptionInfo(False, "Shuffleing tags by "," when create texts."),
     "dataset_filename_word_regex": OptionInfo("", "Filename word regex"),
     "dataset_filename_join_string": OptionInfo(" ", "Filename join string"),
     "training_image_repeats_per_epoch": OptionInfo(1, "Number of repeats for a single input image per epoch; used only for displaying epoch number", gr.Number, {"precision": 0}),
diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index e9d97cc1..df278dc2 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -24,7 +24,7 @@ class DatasetEntry:
 
 
 class PersonalizedBase(Dataset):
-    def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", shuffle_tags=True, model=None, device=None, template_file=None, include_cond=False, batch_size=1):
+    def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", model=None, device=None, template_file=None, include_cond=False, batch_size=1):
         re_word = re.compile(shared.opts.dataset_filename_word_regex) if len(shared.opts.dataset_filename_word_regex) > 0 else None
 
         self.placeholder_token = placeholder_token
@@ -33,7 +33,6 @@ class PersonalizedBase(Dataset):
         self.width = width
         self.height = height
         self.flip = transforms.RandomHorizontalFlip(p=flip_p)
-        self.shuffle_tags = shuffle_tags
 
         self.dataset = []
 
@@ -99,7 +98,7 @@ class PersonalizedBase(Dataset):
     def create_text(self, filename_text):
         text = random.choice(self.lines)
         text = text.replace("[name]", self.placeholder_token)
-        if self.tag_shuffle:
+        if shared.opts.shuffle_tags:
             tags = filename_text.split(',')
             random.shuffle(tags)
             text = text.replace("[filewords]", ','.join(tags))
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index 82dde931..0aeb0459 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -224,7 +224,7 @@ def validate_train_inputs(model_name, learn_rate, batch_size, data_root, templat
     if save_model_every or create_image_every:
         assert log_directory, "Log directory is empty"
 
-def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_directory, training_width, training_height, steps, create_image_every, save_embedding_every, template_file, save_image_with_stored_embedding, preview_from_txt2img, shuffle_tags, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
+def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_directory, training_width, training_height, steps, create_image_every, save_embedding_every, template_file, save_image_with_stored_embedding, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
     save_embedding_every = save_embedding_every or 0
     create_image_every = create_image_every or 0
     validate_train_inputs(embedding_name, learn_rate, batch_size, data_root, template_file, steps, save_embedding_every, create_image_every, log_directory, name="embedding")
@@ -272,7 +272,7 @@ def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_direc
     # dataset loading may take a while, so input validations and early returns should be done before this
     shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..."
     with torch.autocast("cuda"):
-        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=embedding_name, shuffle_tags=shuffle_tags, model=shared.sd_model, device=devices.device, template_file=template_file, batch_size=batch_size)
+        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=embedding_name, model=shared.sd_model, device=devices.device, template_file=template_file, batch_size=batch_size)
     if unload:
         shared.sd_model.first_stage_model.to(devices.cpu)
 
-- 
cgit v1.2.3


From 13a2f1dca32980339e1fb4d1995cde428db798c5 Mon Sep 17 00:00:00 2001
From: KyuSeok Jung <wjdrbtjr495@gmail.com>
Date: Fri, 11 Nov 2022 10:29:55 +0900
Subject: adding tag drop out option

---
 modules/textual_inversion/dataset.py | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index df278dc2..a95c7835 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -98,12 +98,12 @@ class PersonalizedBase(Dataset):
     def create_text(self, filename_text):
         text = random.choice(self.lines)
         text = text.replace("[name]", self.placeholder_token)
+        tags = filename_text.split(',')
+        if shared.opt.tag_drop_out != 0:
+            tags = [t for t in tags if random.random() > shared.opt.tag_drop_out]
         if shared.opts.shuffle_tags:
-            tags = filename_text.split(',')
             random.shuffle(tags)
-            text = text.replace("[filewords]", ','.join(tags))
-        else:
-            text = text.replace("[filewords]", filename_text)
+        text = text.replace("[filewords]", ','.join(tags))
         return text
 
     def __len__(self):
-- 
cgit v1.2.3


From b19af67d29356f97fea5cccfdfa12583f605243f Mon Sep 17 00:00:00 2001
From: KyuSeok Jung <wjdrbtjr495@gmail.com>
Date: Fri, 11 Nov 2022 10:54:19 +0900
Subject: Update dataset.py

---
 modules/textual_inversion/dataset.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index a95c7835..e2cb8428 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -99,7 +99,7 @@ class PersonalizedBase(Dataset):
         text = random.choice(self.lines)
         text = text.replace("[name]", self.placeholder_token)
         tags = filename_text.split(',')
-        if shared.opt.tag_drop_out != 0:
+        if shared.opts.tag_drop_out != 0:
             tags = [t for t in tags if random.random() > shared.opt.tag_drop_out]
         if shared.opts.shuffle_tags:
             random.shuffle(tags)
-- 
cgit v1.2.3


From a1e271207dfc3e89b1286ba41d96b459f210c4b2 Mon Sep 17 00:00:00 2001
From: KyuSeok Jung <wjdrbtjr495@gmail.com>
Date: Fri, 11 Nov 2022 10:56:53 +0900
Subject: Update dataset.py

---
 modules/textual_inversion/dataset.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index e2cb8428..eb75c376 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -100,7 +100,7 @@ class PersonalizedBase(Dataset):
         text = text.replace("[name]", self.placeholder_token)
         tags = filename_text.split(',')
         if shared.opts.tag_drop_out != 0:
-            tags = [t for t in tags if random.random() > shared.opt.tag_drop_out]
+            tags = [t for t in tags if random.random() > shared.opts.tag_drop_out]
         if shared.opts.shuffle_tags:
             random.shuffle(tags)
         text = text.replace("[filewords]", ','.join(tags))
-- 
cgit v1.2.3


From 9a1aff645a4bea745145c57c96950fbd3fcca27c Mon Sep 17 00:00:00 2001
From: parasi <kingofint22@gmail.com>
Date: Sun, 13 Nov 2022 13:44:27 -0600
Subject: resolve [name] after resolving [filewords] in training

---
 modules/textual_inversion/dataset.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index eb75c376..06f271f9 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -97,13 +97,13 @@ class PersonalizedBase(Dataset):
 
     def create_text(self, filename_text):
         text = random.choice(self.lines)
-        text = text.replace("[name]", self.placeholder_token)
         tags = filename_text.split(',')
         if shared.opts.tag_drop_out != 0:
             tags = [t for t in tags if random.random() > shared.opts.tag_drop_out]
         if shared.opts.shuffle_tags:
             random.shuffle(tags)
         text = text.replace("[filewords]", ','.join(tags))
+        text = text.replace("[name]", self.placeholder_token)
         return text
 
     def __len__(self):
-- 
cgit v1.2.3


From bd68e35de3b7cf7547ed97d8bdf60147402133cc Mon Sep 17 00:00:00 2001
From: flamelaw <flamelaw.com3d2@gmail.com>
Date: Sun, 20 Nov 2022 12:35:26 +0900
Subject: Gradient accumulation, autocast fix, new latent sampling method, etc

---
 modules/hypernetworks/hypernetwork.py          | 269 +++++++++++----------
 modules/sd_hijack.py                           |   9 +-
 modules/sd_hijack_checkpoint.py                |  10 +
 modules/shared.py                              |   3 +-
 modules/textual_inversion/dataset.py           | 134 +++++++----
 modules/textual_inversion/textual_inversion.py | 320 ++++++++++++++-----------
 modules/ui.py                                  |  16 +-
 7 files changed, 448 insertions(+), 313 deletions(-)
 create mode 100644 modules/sd_hijack_checkpoint.py

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py
index fbb87dd1..3d3301b0 100644
--- a/modules/hypernetworks/hypernetwork.py
+++ b/modules/hypernetworks/hypernetwork.py
@@ -367,13 +367,13 @@ def report_statistics(loss_info:dict):
 
 
 
-def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log_directory, training_width, training_height, steps, create_image_every, save_hypernetwork_every, template_file, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
+def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, gradient_step, data_root, log_directory, training_width, training_height, steps, shuffle_tags, tag_drop_out, latent_sampling_method, create_image_every, save_hypernetwork_every, template_file, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
     # images allows training previews to have infotext. Importing it at the top causes a circular import problem.
     from modules import images
 
     save_hypernetwork_every = save_hypernetwork_every or 0
     create_image_every = create_image_every or 0
-    textual_inversion.validate_train_inputs(hypernetwork_name, learn_rate, batch_size, data_root, template_file, steps, save_hypernetwork_every, create_image_every, log_directory, name="hypernetwork")
+    textual_inversion.validate_train_inputs(hypernetwork_name, learn_rate, batch_size, gradient_step, data_root, template_file, steps, save_hypernetwork_every, create_image_every, log_directory, name="hypernetwork")
 
     path = shared.hypernetworks.get(hypernetwork_name, None)
     shared.loaded_hypernetwork = Hypernetwork()
@@ -403,28 +403,24 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
     hypernetwork = shared.loaded_hypernetwork
     checkpoint = sd_models.select_checkpoint()
 
-    ititial_step = hypernetwork.step or 0
-    if ititial_step >= steps:
+    initial_step = hypernetwork.step or 0
+    if initial_step >= steps:
         shared.state.textinfo = f"Model has already been trained beyond specified max steps"
         return hypernetwork, filename
 
-    scheduler = LearnRateScheduler(learn_rate, steps, ititial_step)
-    
+    scheduler = LearnRateScheduler(learn_rate, steps, initial_step)
+
     # dataset loading may take a while, so input validations and early returns should be done before this
     shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..."
-    with torch.autocast("cuda"):
-        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=hypernetwork_name, model=shared.sd_model, device=devices.device, template_file=template_file, include_cond=True, batch_size=batch_size)
+
+    pin_memory = shared.opts.pin_memory
+
+    ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=hypernetwork_name, model=shared.sd_model, cond_model=shared.sd_model.cond_stage_model, device=devices.device, template_file=template_file, include_cond=True, batch_size=batch_size, gradient_step=gradient_step, shuffle_tags=shuffle_tags, tag_drop_out=tag_drop_out, latent_sampling_method=latent_sampling_method)
+    dl = modules.textual_inversion.dataset.PersonalizedDataLoader(ds, batch_size=ds.batch_size, pin_memory=pin_memory)
 
     if unload:
         shared.sd_model.cond_stage_model.to(devices.cpu)
         shared.sd_model.first_stage_model.to(devices.cpu)
-
-    size = len(ds.indexes)
-    loss_dict = defaultdict(lambda : deque(maxlen = 1024))
-    losses = torch.zeros((size,))
-    previous_mean_losses = [0]
-    previous_mean_loss = 0
-    print("Mean loss of {} elements".format(size))
     
     weights = hypernetwork.weights()
     for weight in weights:
@@ -436,8 +432,8 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
         optimizer_name = hypernetwork.optimizer_name
     else:
         print(f"Optimizer type {hypernetwork.optimizer_name} is not defined!")
-        optimizer = torch.optim.AdamW(params=weights, lr=scheduler.learn_rate)
-        optimizer_name = 'AdamW'
+    optimizer = torch.optim.AdamW(params=weights, lr=scheduler.learn_rate)
+    optimizer_name = 'AdamW'
 
     if hypernetwork.optimizer_state_dict:  # This line must be changed if Optimizer type can be different from saved optimizer.
         try:
@@ -446,131 +442,155 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
             print("Cannot resume from saved optimizer!")
             print(e)
 
+    scaler = torch.cuda.amp.GradScaler()
+    
+    batch_size = ds.batch_size
+    gradient_step = ds.gradient_step
+    # n steps = batch_size * gradient_step * n image processed
+    steps_per_epoch = len(ds) // batch_size // gradient_step
+    max_steps_per_epoch = len(ds) // batch_size - (len(ds) // batch_size) % gradient_step
+    loss_step = 0
+    _loss_step = 0 #internal
+    # size = len(ds.indexes)
+    # loss_dict = defaultdict(lambda : deque(maxlen = 1024))
+    # losses = torch.zeros((size,))
+    # previous_mean_losses = [0]
+    # previous_mean_loss = 0
+    # print("Mean loss of {} elements".format(size))
+
     steps_without_grad = 0
 
     last_saved_file = "<none>"
     last_saved_image = "<none>"
     forced_filename = "<none>"
 
-    pbar = tqdm.tqdm(enumerate(ds), total=steps - ititial_step)
-    for i, entries in pbar:
-        hypernetwork.step = i + ititial_step
-        if len(loss_dict) > 0:
-            previous_mean_losses = [i[-1] for i in loss_dict.values()]
-            previous_mean_loss = mean(previous_mean_losses)
-            
-        scheduler.apply(optimizer, hypernetwork.step)
-        if scheduler.finished:
-            break
-
-        if shared.state.interrupted:
-            break
-
-        with torch.autocast("cuda"):
-            c = stack_conds([entry.cond for entry in entries]).to(devices.device)
-            # c = torch.vstack([entry.cond for entry in entries]).to(devices.device)
-            x = torch.stack([entry.latent for entry in entries]).to(devices.device)
-            loss = shared.sd_model(x, c)[0]
-            del x
-            del c
-
-            losses[hypernetwork.step % losses.shape[0]] = loss.item()
-            for entry in entries:
-                loss_dict[entry.filename].append(loss.item())
+    pbar = tqdm.tqdm(total=steps - initial_step)
+    try:
+        for i in range((steps-initial_step) * gradient_step):
+            if scheduler.finished:
+                break
+            if shared.state.interrupted:
+                break
+            for j, batch in enumerate(dl):
+                # works as a drop_last=True for gradient accumulation
+                if j == max_steps_per_epoch:
+                    break
+                scheduler.apply(optimizer, hypernetwork.step)
+                if scheduler.finished:
+                    break
+                if shared.state.interrupted:
+                    break
+
+                with torch.autocast("cuda"):
+                    x = batch.latent_sample.to(devices.device, non_blocking=pin_memory)
+                    if tag_drop_out != 0 or shuffle_tags:
+                        shared.sd_model.cond_stage_model.to(devices.device)
+                        c = shared.sd_model.cond_stage_model(batch.cond_text).to(devices.device, non_blocking=pin_memory)
+                        shared.sd_model.cond_stage_model.to(devices.cpu)
+                    else:
+                        c = stack_conds(batch.cond).to(devices.device, non_blocking=pin_memory)
+                    loss = shared.sd_model(x, c)[0] / gradient_step
+                    del x
+                    del c
+
+                    _loss_step += loss.item()
+                scaler.scale(loss).backward()
+                # go back until we reach gradient accumulation steps
+                if (j + 1) % gradient_step != 0:
+                    continue
+                # print(f"grad:{weights[0].grad.detach().cpu().abs().mean().item():.7f}")
+                # scaler.unscale_(optimizer)
+                # print(f"grad:{weights[0].grad.detach().cpu().abs().mean().item():.15f}")
+                # torch.nn.utils.clip_grad_norm_(weights, max_norm=1.0)
+                # print(f"grad:{weights[0].grad.detach().cpu().abs().mean().item():.15f}")
+                scaler.step(optimizer)
+                scaler.update()
+                hypernetwork.step += 1
+                pbar.update()
+                optimizer.zero_grad(set_to_none=True)
+                loss_step = _loss_step
+                _loss_step = 0
+
+                steps_done = hypernetwork.step + 1
                 
-            optimizer.zero_grad()
-            weights[0].grad = None
-            loss.backward()
-
-            if weights[0].grad is None:
-                steps_without_grad += 1
-            else:
-                steps_without_grad = 0
-            assert steps_without_grad < 10, 'no gradient found for the trained weight after backward() for 10 steps in a row; this is a bug; training cannot continue'
-
-            optimizer.step()
-
-        steps_done = hypernetwork.step + 1
-
-        if torch.isnan(losses[hypernetwork.step % losses.shape[0]]): 
-            raise RuntimeError("Loss diverged.")
-        
-        if len(previous_mean_losses) > 1:
-            std = stdev(previous_mean_losses)
-        else:
-            std = 0
-        dataset_loss_info = f"dataset loss:{mean(previous_mean_losses):.3f}" + u"\u00B1" + f"({std / (len(previous_mean_losses) ** 0.5):.3f})"
-        pbar.set_description(dataset_loss_info)
-
-        if hypernetwork_dir is not None and steps_done % save_hypernetwork_every == 0:
-            # Before saving, change name to match current checkpoint.
-            hypernetwork_name_every = f'{hypernetwork_name}-{steps_done}'
-            last_saved_file = os.path.join(hypernetwork_dir, f'{hypernetwork_name_every}.pt')
-            hypernetwork.optimizer_name = optimizer_name
-            if shared.opts.save_optimizer_state:
-                hypernetwork.optimizer_state_dict = optimizer.state_dict()
-            save_hypernetwork(hypernetwork, checkpoint, hypernetwork_name, last_saved_file)
-            hypernetwork.optimizer_state_dict = None  # dereference it after saving, to save memory.
-
-        textual_inversion.write_loss(log_directory, "hypernetwork_loss.csv", hypernetwork.step, len(ds), {
-            "loss": f"{previous_mean_loss:.7f}",
-            "learn_rate": scheduler.learn_rate
-        })
-
-        if images_dir is not None and steps_done % create_image_every == 0:
-            forced_filename = f'{hypernetwork_name}-{steps_done}'
-            last_saved_image = os.path.join(images_dir, forced_filename)
-
-            optimizer.zero_grad()
-            shared.sd_model.cond_stage_model.to(devices.device)
-            shared.sd_model.first_stage_model.to(devices.device)
-
-            p = processing.StableDiffusionProcessingTxt2Img(
-                sd_model=shared.sd_model,
-                do_not_save_grid=True,
-                do_not_save_samples=True,
-            )
-
-            if preview_from_txt2img:
-                p.prompt = preview_prompt
-                p.negative_prompt = preview_negative_prompt
-                p.steps = preview_steps
-                p.sampler_name = sd_samplers.samplers[preview_sampler_index].name
-                p.cfg_scale = preview_cfg_scale
-                p.seed = preview_seed
-                p.width = preview_width
-                p.height = preview_height
-            else:
-                p.prompt = entries[0].cond_text
-                p.steps = 20
+                epoch_num = hypernetwork.step // steps_per_epoch
+                epoch_step = hypernetwork.step % steps_per_epoch
+
+                pbar.set_description(f"[Epoch {epoch_num}: {epoch_step+1}/{steps_per_epoch}]loss: {loss_step:.7f}")
+                if hypernetwork_dir is not None and steps_done % save_hypernetwork_every == 0:
+                    # Before saving, change name to match current checkpoint.
+                    hypernetwork_name_every = f'{hypernetwork_name}-{steps_done}'
+                    last_saved_file = os.path.join(hypernetwork_dir, f'{hypernetwork_name_every}.pt')
+                    hypernetwork.optimizer_name = optimizer_name
+                    if shared.opts.save_optimizer_state:
+                        hypernetwork.optimizer_state_dict = optimizer.state_dict()
+                    save_hypernetwork(hypernetwork, checkpoint, hypernetwork_name, last_saved_file)
+                    hypernetwork.optimizer_state_dict = None  # dereference it after saving, to save memory.
+
+                textual_inversion.write_loss(log_directory, "hypernetwork_loss.csv", hypernetwork.step, steps_per_epoch, {
+                    "loss": f"{loss_step:.7f}",
+                    "learn_rate": scheduler.learn_rate
+                })
+
+                if images_dir is not None and steps_done % create_image_every == 0:
+                    forced_filename = f'{hypernetwork_name}-{steps_done}'
+                    last_saved_image = os.path.join(images_dir, forced_filename)
+
+                    shared.sd_model.cond_stage_model.to(devices.device)
+                    shared.sd_model.first_stage_model.to(devices.device)
+
+                    p = processing.StableDiffusionProcessingTxt2Img(
+                        sd_model=shared.sd_model,
+                        do_not_save_grid=True,
+                        do_not_save_samples=True,
+                    )
+
+                    if preview_from_txt2img:
+                        p.prompt = preview_prompt
+                        p.negative_prompt = preview_negative_prompt
+                        p.steps = preview_steps
+                        p.sampler_name = sd_samplers.samplers[preview_sampler_index].name
+                        p.cfg_scale = preview_cfg_scale
+                        p.seed = preview_seed
+                        p.width = preview_width
+                        p.height = preview_height
+                    else:
+                        p.prompt = batch.cond_text[0]
+                        p.steps = 20
+                        p.width = training_width
+                        p.height = training_height
 
-            preview_text = p.prompt
+                    preview_text = p.prompt
 
-            processed = processing.process_images(p)
-            image = processed.images[0] if len(processed.images)>0 else None
+                    processed = processing.process_images(p)
+                    image = processed.images[0] if len(processed.images) > 0 else None
 
-            if unload:
-                shared.sd_model.cond_stage_model.to(devices.cpu)
-                shared.sd_model.first_stage_model.to(devices.cpu)
+                    if unload:
+                        shared.sd_model.cond_stage_model.to(devices.cpu)
+                        shared.sd_model.first_stage_model.to(devices.cpu)
 
-            if image is not None:
-                shared.state.current_image = image
-                last_saved_image, last_text_info = images.save_image(image, images_dir, "", p.seed, p.prompt, shared.opts.samples_format, processed.infotexts[0], p=p, forced_filename=forced_filename, save_to_dirs=False)
-                last_saved_image += f", prompt: {preview_text}"
+                    if image is not None:
+                        shared.state.current_image = image
+                        last_saved_image, last_text_info = images.save_image(image, images_dir, "", p.seed, p.prompt, shared.opts.samples_format, processed.infotexts[0], p=p, forced_filename=forced_filename, save_to_dirs=False)
+                        last_saved_image += f", prompt: {preview_text}"
 
-        shared.state.job_no = hypernetwork.step
+                shared.state.job_no = hypernetwork.step
 
-        shared.state.textinfo = f"""
+                shared.state.textinfo = f"""
 <p>
-Loss: {previous_mean_loss:.7f}<br/>
+Loss: {loss_step:.7f}<br/>
 Step: {hypernetwork.step}<br/>
-Last prompt: {html.escape(entries[0].cond_text)}<br/>
+Last prompt: {html.escape(batch.cond_text[0])}<br/>
 Last saved hypernetwork: {html.escape(last_saved_file)}<br/>
 Last saved image: {html.escape(last_saved_image)}<br/>
 </p>
 """
-        
-    report_statistics(loss_dict)
+    except Exception:
+        print(traceback.format_exc(), file=sys.stderr)
+    finally:
+        pbar.leave = False
+        pbar.close()
+        #report_statistics(loss_dict)
 
     filename = os.path.join(shared.cmd_opts.hypernetwork_dir, f'{hypernetwork_name}.pt')
     hypernetwork.optimizer_name = optimizer_name
@@ -579,6 +599,9 @@ Last saved image: {html.escape(last_saved_image)}<br/>
     save_hypernetwork(hypernetwork, checkpoint, hypernetwork_name, filename)
     del optimizer
     hypernetwork.optimizer_state_dict = None  # dereference it after saving, to save memory.
+    shared.sd_model.cond_stage_model.to(devices.device)
+    shared.sd_model.first_stage_model.to(devices.device)
+
     return hypernetwork, filename
 
 def save_hypernetwork(hypernetwork, checkpoint, hypernetwork_name, filename):
diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py
index eaedac13..29c8b561 100644
--- a/modules/sd_hijack.py
+++ b/modules/sd_hijack.py
@@ -8,7 +8,7 @@ from torch import einsum
 from torch.nn.functional import silu
 
 import modules.textual_inversion.textual_inversion
-from modules import prompt_parser, devices, sd_hijack_optimizations, shared
+from modules import prompt_parser, devices, sd_hijack_optimizations, shared, sd_hijack_checkpoint
 from modules.shared import opts, device, cmd_opts
 from modules.sd_hijack_optimizations import invokeAI_mps_available
 
@@ -59,6 +59,10 @@ def undo_optimizations():
 def get_target_prompt_token_count(token_count):
     return math.ceil(max(token_count, 1) / 75) * 75
 
+def fix_checkpoint():
+    ldm.modules.attention.BasicTransformerBlock.forward = sd_hijack_checkpoint.BasicTransformerBlock_forward
+    ldm.modules.diffusionmodules.openaimodel.ResBlock.forward = sd_hijack_checkpoint.ResBlock_forward
+    ldm.modules.diffusionmodules.openaimodel.AttentionBlock.forward = sd_hijack_checkpoint.AttentionBlock_forward
 
 class StableDiffusionModelHijack:
     fixes = None
@@ -78,6 +82,7 @@ class StableDiffusionModelHijack:
         self.clip = m.cond_stage_model
 
         apply_optimizations()
+        fix_checkpoint()
 
         def flatten(el):
             flattened = [flatten(children) for children in el.children()]
@@ -303,7 +308,7 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module):
             batch_multipliers, remade_batch_tokens, used_custom_terms, hijack_comments, hijack_fixes, token_count = self.process_text_old(text)
         else:
             batch_multipliers, remade_batch_tokens, used_custom_terms, hijack_comments, hijack_fixes, token_count = self.process_text(text)
-
+        
         self.hijack.comments += hijack_comments
 
         if len(used_custom_terms) > 0:
diff --git a/modules/sd_hijack_checkpoint.py b/modules/sd_hijack_checkpoint.py
new file mode 100644
index 00000000..5712972f
--- /dev/null
+++ b/modules/sd_hijack_checkpoint.py
@@ -0,0 +1,10 @@
+from torch.utils.checkpoint import checkpoint
+
+def BasicTransformerBlock_forward(self, x, context=None):
+    return checkpoint(self._forward, x, context)
+
+def AttentionBlock_forward(self, x):
+    return checkpoint(self._forward, x)
+
+def ResBlock_forward(self, x, emb):
+    return checkpoint(self._forward, x, emb)
\ No newline at end of file
diff --git a/modules/shared.py b/modules/shared.py
index a4457305..3704ce23 100644
--- a/modules/shared.py
+++ b/modules/shared.py
@@ -322,8 +322,7 @@ options_templates.update(options_section(('system', "System"), {
 
 options_templates.update(options_section(('training', "Training"), {
     "unload_models_when_training": OptionInfo(False, "Move VAE and CLIP to RAM when training if possible. Saves VRAM."),
-    "shuffle_tags": OptionInfo(False, "Shuffleing tags by ',' when create texts."),
-    "tag_drop_out": OptionInfo(0, "Dropout tags when create texts", gr.Slider, {"minimum": 0, "maximum": 1, "step": 0.1}),
+    "pin_memory": OptionInfo(False, "Turn on pin_memory for DataLoader. Makes training slightly faster but can increase memory usage."),
     "save_optimizer_state": OptionInfo(False, "Saves Optimizer state as separate *.optim file. Training can be resumed with HN itself and matching optim file."),
     "dataset_filename_word_regex": OptionInfo("", "Filename word regex"),
     "dataset_filename_join_string": OptionInfo(" ", "Filename join string"),
diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index eb75c376..d594b49d 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -3,7 +3,7 @@ import numpy as np
 import PIL
 import torch
 from PIL import Image
-from torch.utils.data import Dataset
+from torch.utils.data import Dataset, DataLoader
 from torchvision import transforms
 
 import random
@@ -11,25 +11,28 @@ import tqdm
 from modules import devices, shared
 import re
 
+from ldm.modules.distributions.distributions import DiagonalGaussianDistribution
+
 re_numbers_at_start = re.compile(r"^[-\d]+\s*")
 
 
 class DatasetEntry:
-    def __init__(self, filename=None, latent=None, filename_text=None):
+    def __init__(self, filename=None, filename_text=None, latent_dist=None, latent_sample=None, cond=None, cond_text=None, pixel_values=None):
         self.filename = filename
-        self.latent = latent
         self.filename_text = filename_text
-        self.cond = None
-        self.cond_text = None
+        self.latent_dist = latent_dist
+        self.latent_sample = latent_sample
+        self.cond = cond
+        self.cond_text = cond_text
+        self.pixel_values = pixel_values
 
 
 class PersonalizedBase(Dataset):
-    def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", model=None, device=None, template_file=None, include_cond=False, batch_size=1):
+    def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", model=None, cond_model=None, device=None, template_file=None, include_cond=False, batch_size=1, gradient_step=1, shuffle_tags=False, tag_drop_out=0, latent_sampling_method='once'):        
         re_word = re.compile(shared.opts.dataset_filename_word_regex) if len(shared.opts.dataset_filename_word_regex) > 0 else None
-
+        
         self.placeholder_token = placeholder_token
 
-        self.batch_size = batch_size
         self.width = width
         self.height = height
         self.flip = transforms.RandomHorizontalFlip(p=flip_p)
@@ -45,11 +48,16 @@ class PersonalizedBase(Dataset):
         assert os.path.isdir(data_root), "Dataset directory doesn't exist"
         assert os.listdir(data_root), "Dataset directory is empty"
 
-        cond_model = shared.sd_model.cond_stage_model
-
         self.image_paths = [os.path.join(data_root, file_path) for file_path in os.listdir(data_root)]
+
+        
+        self.shuffle_tags = shuffle_tags
+        self.tag_drop_out = tag_drop_out
+
         print("Preparing dataset...")
         for path in tqdm.tqdm(self.image_paths):
+            if shared.state.interrupted:
+                raise Exception("inturrupted")
             try:
                 image = Image.open(path).convert('RGB').resize((self.width, self.height), PIL.Image.BICUBIC)
             except Exception:
@@ -71,37 +79,58 @@ class PersonalizedBase(Dataset):
             npimage = np.array(image).astype(np.uint8)
             npimage = (npimage / 127.5 - 1.0).astype(np.float32)
 
-            torchdata = torch.from_numpy(npimage).to(device=device, dtype=torch.float32)
-            torchdata = torch.moveaxis(torchdata, 2, 0)
-
-            init_latent = model.get_first_stage_encoding(model.encode_first_stage(torchdata.unsqueeze(dim=0))).squeeze()
-            init_latent = init_latent.to(devices.cpu)
-
-            entry = DatasetEntry(filename=path, filename_text=filename_text, latent=init_latent)
-
-            if include_cond:
+            torchdata = torch.from_numpy(npimage).permute(2, 0, 1).to(device=device, dtype=torch.float32)
+            latent_sample = None
+
+            with torch.autocast("cuda"):
+                latent_dist = model.encode_first_stage(torchdata.unsqueeze(dim=0))
+
+            if latent_sampling_method == "once" or (latent_sampling_method == "deterministic" and not isinstance(latent_dist, DiagonalGaussianDistribution)):
+                latent_sample = model.get_first_stage_encoding(latent_dist).squeeze().to(devices.cpu)
+                latent_sampling_method = "once"
+                entry = DatasetEntry(filename=path, filename_text=filename_text, latent_sample=latent_sample)
+            elif latent_sampling_method == "deterministic":
+                # Works only for DiagonalGaussianDistribution
+                latent_dist.std = 0
+                latent_sample = model.get_first_stage_encoding(latent_dist).squeeze().to(devices.cpu)
+                entry = DatasetEntry(filename=path, filename_text=filename_text, latent_sample=latent_sample)
+            elif latent_sampling_method == "random":
+                entry = DatasetEntry(filename=path, filename_text=filename_text, latent_dist=latent_dist)
+
+            if not (self.tag_drop_out != 0 or self.shuffle_tags):
                 entry.cond_text = self.create_text(filename_text)
-                entry.cond = cond_model([entry.cond_text]).to(devices.cpu).squeeze(0)
 
-            self.dataset.append(entry)
-
-        assert len(self.dataset) > 0, "No images have been found in the dataset."
-        self.length = len(self.dataset) * repeats // batch_size
+            if include_cond and not (self.tag_drop_out != 0 or self.shuffle_tags):
+                with torch.autocast("cuda"):
+                    entry.cond = cond_model([entry.cond_text]).to(devices.cpu).squeeze(0)
+            # elif not include_cond:
+            #     _, _, _, _, hijack_fixes, token_count = cond_model.process_text([entry.cond_text])
+            #     max_n = token_count // 75
+            #     index_list = [ [] for _ in range(max_n + 1) ]
+            #     for n, (z, _) in hijack_fixes[0]:
+            #         index_list[n].append(z)
+            #     with torch.autocast("cuda"):
+            #         entry.cond = cond_model([entry.cond_text]).to(devices.cpu).squeeze(0)
+            #     entry.emb_index = index_list
 
-        self.dataset_length = len(self.dataset)
-        self.indexes = None
-        self.shuffle()
+            self.dataset.append(entry)
+            del torchdata
+            del latent_dist
+            del latent_sample
 
-    def shuffle(self):
-        self.indexes = np.random.permutation(self.dataset_length)
+        self.length = len(self.dataset)
+        assert self.length > 0, "No images have been found in the dataset."
+        self.batch_size = min(batch_size, self.length)
+        self.gradient_step = min(gradient_step, self.length // self.batch_size)
+        self.latent_sampling_method = latent_sampling_method
 
     def create_text(self, filename_text):
         text = random.choice(self.lines)
         text = text.replace("[name]", self.placeholder_token)
         tags = filename_text.split(',')
-        if shared.opts.tag_drop_out != 0:
-            tags = [t for t in tags if random.random() > shared.opts.tag_drop_out]
-        if shared.opts.shuffle_tags:
+        if self.tag_drop_out != 0:
+            tags = [t for t in tags if random.random() > self.tag_drop_out]
+        if self.shuffle_tags:
             random.shuffle(tags)
         text = text.replace("[filewords]", ','.join(tags))
         return text
@@ -110,19 +139,28 @@ class PersonalizedBase(Dataset):
         return self.length
 
     def __getitem__(self, i):
-        res = []
-
-        for j in range(self.batch_size):
-            position = i * self.batch_size + j
-            if position % len(self.indexes) == 0:
-                self.shuffle()
-
-            index = self.indexes[position % len(self.indexes)]
-            entry = self.dataset[index]
-
-            if entry.cond is None:
-                entry.cond_text = self.create_text(entry.filename_text)
-
-            res.append(entry)
-
-        return res
+        entry = self.dataset[i]
+        if self.tag_drop_out != 0 or self.shuffle_tags:
+            entry.cond_text = self.create_text(entry.filename_text)
+        if self.latent_sampling_method == "random":
+            entry.latent_sample = shared.sd_model.get_first_stage_encoding(entry.latent_dist)
+        return entry
+
+class PersonalizedDataLoader(DataLoader):
+    def __init__(self, *args, **kwargs):
+        super(PersonalizedDataLoader, self).__init__(shuffle=True, drop_last=True, *args, **kwargs)
+        self.collate_fn = collate_wrapper
+        
+
+class BatchLoader:
+    def __init__(self, data):
+        self.cond_text = [entry.cond_text for entry in data]
+        self.cond = [entry.cond for entry in data]
+        self.latent_sample = torch.stack([entry.latent_sample for entry in data]).squeeze(1)
+
+    def pin_memory(self):
+        self.latent_sample = self.latent_sample.pin_memory()
+        return self
+
+def collate_wrapper(batch):
+    return BatchLoader(batch)
\ No newline at end of file
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index 5e4d8688..1d5e3a32 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -184,7 +184,7 @@ def write_loss(log_directory, filename, step, epoch_len, values):
     if shared.opts.training_write_csv_every == 0:
         return
 
-    if (step + 1) % shared.opts.training_write_csv_every != 0:
+    if step % shared.opts.training_write_csv_every != 0:
         return
     write_csv_header = False if os.path.exists(os.path.join(log_directory, filename)) else True
 
@@ -194,21 +194,23 @@ def write_loss(log_directory, filename, step, epoch_len, values):
         if write_csv_header:
             csv_writer.writeheader()
 
-        epoch = step // epoch_len
-        epoch_step = step % epoch_len 
+        epoch = (step - 1) // epoch_len
+        epoch_step = (step - 1) % epoch_len 
 
         csv_writer.writerow({
-            "step": step + 1,
+            "step": step,
             "epoch": epoch,
-            "epoch_step": epoch_step + 1,
+            "epoch_step": epoch_step,
             **values,
         })
 
-def validate_train_inputs(model_name, learn_rate, batch_size, data_root, template_file, steps, save_model_every, create_image_every, log_directory, name="embedding"):
+def validate_train_inputs(model_name, learn_rate, batch_size, gradient_step, data_root, template_file, steps, save_model_every, create_image_every, log_directory, name="embedding"):
     assert model_name, f"{name} not selected"
     assert learn_rate, "Learning rate is empty or 0"
     assert isinstance(batch_size, int), "Batch size must be integer"
     assert batch_size > 0, "Batch size must be positive"
+    assert isinstance(gradient_step, int), "Gradient accumulation step must be integer"
+    assert gradient_step > 0, "Gradient accumulation step must be positive"
     assert data_root, "Dataset directory is empty"
     assert os.path.isdir(data_root), "Dataset directory doesn't exist"
     assert os.listdir(data_root), "Dataset directory is empty"
@@ -224,10 +226,10 @@ def validate_train_inputs(model_name, learn_rate, batch_size, data_root, templat
     if save_model_every or create_image_every:
         assert log_directory, "Log directory is empty"
 
-def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_directory, training_width, training_height, steps, create_image_every, save_embedding_every, template_file, save_image_with_stored_embedding, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
+def train_embedding(embedding_name, learn_rate, batch_size, gradient_step, data_root, log_directory, training_width, training_height, steps, shuffle_tags, tag_drop_out, latent_sampling_method, create_image_every, save_embedding_every, template_file, save_image_with_stored_embedding, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
     save_embedding_every = save_embedding_every or 0
     create_image_every = create_image_every or 0
-    validate_train_inputs(embedding_name, learn_rate, batch_size, data_root, template_file, steps, save_embedding_every, create_image_every, log_directory, name="embedding")
+    validate_train_inputs(embedding_name, learn_rate, batch_size, gradient_step, data_root, template_file, steps, save_embedding_every, create_image_every, log_directory, name="embedding")
 
     shared.state.textinfo = "Initializing textual inversion training..."
     shared.state.job_count = steps
@@ -255,161 +257,205 @@ def train_embedding(embedding_name, learn_rate, batch_size, data_root, log_direc
     else:
         images_embeds_dir = None
 
-    cond_model = shared.sd_model.cond_stage_model
-
     hijack = sd_hijack.model_hijack
 
     embedding = hijack.embedding_db.word_embeddings[embedding_name]
     checkpoint = sd_models.select_checkpoint()
 
-    ititial_step = embedding.step or 0
-    if ititial_step >= steps:
+    initial_step = embedding.step or 0
+    if initial_step >= steps:
         shared.state.textinfo = f"Model has already been trained beyond specified max steps"
         return embedding, filename
+    scheduler = LearnRateScheduler(learn_rate, steps, initial_step)
 
-    scheduler = LearnRateScheduler(learn_rate, steps, ititial_step)
-
-    # dataset loading may take a while, so input validations and early returns should be done before this
+   # dataset loading may take a while, so input validations and early returns should be done before this
     shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..."
-    with torch.autocast("cuda"):
-        ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=embedding_name, model=shared.sd_model, device=devices.device, template_file=template_file, batch_size=batch_size)
+    
+    pin_memory = shared.opts.pin_memory
+    
+    ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=embedding_name, model=shared.sd_model, cond_model=shared.sd_model.cond_stage_model, device=devices.device, template_file=template_file, batch_size=batch_size, gradient_step=gradient_step, shuffle_tags=shuffle_tags, tag_drop_out=tag_drop_out, latent_sampling_method=latent_sampling_method)
+
+    latent_sampling_method = ds.latent_sampling_method
+
+    dl = modules.textual_inversion.dataset.PersonalizedDataLoader(ds, batch_size=ds.batch_size, pin_memory=False)
+
     if unload:
         shared.sd_model.first_stage_model.to(devices.cpu)
 
     embedding.vec.requires_grad = True
     optimizer = torch.optim.AdamW([embedding.vec], lr=scheduler.learn_rate)
+    scaler = torch.cuda.amp.GradScaler()
 
-    losses = torch.zeros((32,))
+    batch_size = ds.batch_size
+    gradient_step = ds.gradient_step
+    # n steps = batch_size * gradient_step * n image processed
+    steps_per_epoch = len(ds) // batch_size // gradient_step
+    max_steps_per_epoch = len(ds) // batch_size - (len(ds) // batch_size) % gradient_step
+    loss_step = 0
+    _loss_step = 0 #internal
 
+    
     last_saved_file = "<none>"
     last_saved_image = "<none>"
     forced_filename = "<none>"
     embedding_yet_to_be_embedded = False
-
-    pbar = tqdm.tqdm(enumerate(ds), total=steps-ititial_step)
-    for i, entries in pbar:
-        embedding.step = i + ititial_step
-
-        scheduler.apply(optimizer, embedding.step)
-        if scheduler.finished:
-            break
-
-        if shared.state.interrupted:
-            break
-
-        with torch.autocast("cuda"):
-            c = cond_model([entry.cond_text for entry in entries])
-            x = torch.stack([entry.latent for entry in entries]).to(devices.device)
-            loss = shared.sd_model(x, c)[0]
-            del x
-
-            losses[embedding.step % losses.shape[0]] = loss.item()
-
-            optimizer.zero_grad()
-            loss.backward()
-            optimizer.step()
-
-        steps_done = embedding.step + 1
-
-        epoch_num = embedding.step // len(ds)
-        epoch_step = embedding.step % len(ds)
-
-        pbar.set_description(f"[Epoch {epoch_num}: {epoch_step+1}/{len(ds)}]loss: {losses.mean():.7f}")
-
-        if embedding_dir is not None and steps_done % save_embedding_every == 0:
-            # Before saving, change name to match current checkpoint.
-            embedding_name_every = f'{embedding_name}-{steps_done}'
-            last_saved_file = os.path.join(embedding_dir, f'{embedding_name_every}.pt')
-            save_embedding(embedding, checkpoint, embedding_name_every, last_saved_file, remove_cached_checksum=True)
-            embedding_yet_to_be_embedded = True
-
-        write_loss(log_directory, "textual_inversion_loss.csv", embedding.step, len(ds), {
-            "loss": f"{losses.mean():.7f}",
-            "learn_rate": scheduler.learn_rate
-        })
-
-        if images_dir is not None and steps_done % create_image_every == 0:
-            forced_filename = f'{embedding_name}-{steps_done}'
-            last_saved_image = os.path.join(images_dir, forced_filename)
-
-            shared.sd_model.first_stage_model.to(devices.device)
-
-            p = processing.StableDiffusionProcessingTxt2Img(
-                sd_model=shared.sd_model,
-                do_not_save_grid=True,
-                do_not_save_samples=True,
-                do_not_reload_embeddings=True,
-            )
-
-            if preview_from_txt2img:
-                p.prompt = preview_prompt
-                p.negative_prompt = preview_negative_prompt
-                p.steps = preview_steps
-                p.sampler_name = sd_samplers.samplers[preview_sampler_index].name
-                p.cfg_scale = preview_cfg_scale
-                p.seed = preview_seed
-                p.width = preview_width
-                p.height = preview_height
-            else:
-                p.prompt = entries[0].cond_text
-                p.steps = 20
-                p.width = training_width
-                p.height = training_height
-
-            preview_text = p.prompt
-
-            processed = processing.process_images(p)
-            image = processed.images[0]
-
-            if unload:
-                shared.sd_model.first_stage_model.to(devices.cpu)
-
-            shared.state.current_image = image
-
-            if save_image_with_stored_embedding and os.path.exists(last_saved_file) and embedding_yet_to_be_embedded:
-
-                last_saved_image_chunks = os.path.join(images_embeds_dir, f'{embedding_name}-{steps_done}.png')
-
-                info = PngImagePlugin.PngInfo()
-                data = torch.load(last_saved_file)
-                info.add_text("sd-ti-embedding", embedding_to_b64(data))
-
-                title = "<{}>".format(data.get('name', '???'))
-
-                try:
-                    vectorSize = list(data['string_to_param'].values())[0].shape[0]
-                except Exception as e:
-                    vectorSize = '?'
-
-                checkpoint = sd_models.select_checkpoint()
-                footer_left = checkpoint.model_name
-                footer_mid = '[{}]'.format(checkpoint.hash)
-                footer_right = '{}v {}s'.format(vectorSize, steps_done)
-
-                captioned_image = caption_image_overlay(image, title, footer_left, footer_mid, footer_right)
-                captioned_image = insert_image_data_embed(captioned_image, data)
-
-                captioned_image.save(last_saved_image_chunks, "PNG", pnginfo=info)
-                embedding_yet_to_be_embedded = False
-
-            last_saved_image, last_text_info = images.save_image(image, images_dir, "", p.seed, p.prompt, shared.opts.samples_format, processed.infotexts[0], p=p, forced_filename=forced_filename, save_to_dirs=False)
-            last_saved_image += f", prompt: {preview_text}"
-
-        shared.state.job_no = embedding.step
-
-        shared.state.textinfo = f"""
+    
+    pbar = tqdm.tqdm(total=steps - initial_step)
+    try:
+        for i in range((steps-initial_step) * gradient_step):
+            if scheduler.finished:
+                break
+            if shared.state.interrupted:
+                break
+            for j, batch in enumerate(dl):
+                # works as a drop_last=True for gradient accumulation
+                if j == max_steps_per_epoch:
+                    break
+                scheduler.apply(optimizer, embedding.step)
+                if scheduler.finished:
+                    break
+                if shared.state.interrupted:
+                    break
+
+                with torch.autocast("cuda"):
+                    # c = stack_conds(batch.cond).to(devices.device)
+                    # mask = torch.tensor(batch.emb_index).to(devices.device, non_blocking=pin_memory)
+                    # print(mask)
+                    # c[:, 1:1+embedding.vec.shape[0]] = embedding.vec.to(devices.device, non_blocking=pin_memory)
+                    x = batch.latent_sample.to(devices.device, non_blocking=pin_memory)
+                    c = shared.sd_model.cond_stage_model(batch.cond_text)
+                    loss = shared.sd_model(x, c)[0] / gradient_step
+                    del x
+                    
+                    _loss_step += loss.item()
+                scaler.scale(loss).backward()
+                
+                # go back until we reach gradient accumulation steps
+                if (j + 1) % gradient_step != 0:
+                    continue
+                #print(f"grad:{embedding.vec.grad.detach().cpu().abs().mean().item():.7f}")
+                #scaler.unscale_(optimizer)
+                #print(f"grad:{embedding.vec.grad.detach().cpu().abs().mean().item():.7f}")
+                #torch.nn.utils.clip_grad_norm_(embedding.vec, max_norm=1.0)
+                #print(f"grad:{embedding.vec.grad.detach().cpu().abs().mean().item():.7f}")
+                scaler.step(optimizer)
+                scaler.update()
+                embedding.step += 1
+                pbar.update()
+                optimizer.zero_grad(set_to_none=True)
+                loss_step = _loss_step
+                _loss_step = 0
+
+                steps_done = embedding.step + 1
+
+                epoch_num = embedding.step // steps_per_epoch
+                epoch_step = embedding.step % steps_per_epoch
+
+                pbar.set_description(f"[Epoch {epoch_num}: {epoch_step+1}/{steps_per_epoch}]loss: {loss_step:.7f}")
+                if embedding_dir is not None and steps_done % save_embedding_every == 0:
+                    # Before saving, change name to match current checkpoint.
+                    embedding_name_every = f'{embedding_name}-{steps_done}'
+                    last_saved_file = os.path.join(embedding_dir, f'{embedding_name_every}.pt')
+                    #if shared.opts.save_optimizer_state:
+                        #embedding.optimizer_state_dict = optimizer.state_dict()
+                    save_embedding(embedding, checkpoint, embedding_name_every, last_saved_file, remove_cached_checksum=True)
+                    embedding_yet_to_be_embedded = True
+
+                write_loss(log_directory, "textual_inversion_loss.csv", embedding.step, steps_per_epoch, {
+                    "loss": f"{loss_step:.7f}",
+                    "learn_rate": scheduler.learn_rate
+                })
+
+                if images_dir is not None and steps_done % create_image_every == 0:
+                    forced_filename = f'{embedding_name}-{steps_done}'
+                    last_saved_image = os.path.join(images_dir, forced_filename)
+
+                    shared.sd_model.first_stage_model.to(devices.device)
+
+                    p = processing.StableDiffusionProcessingTxt2Img(
+                        sd_model=shared.sd_model,
+                        do_not_save_grid=True,
+                        do_not_save_samples=True,
+                        do_not_reload_embeddings=True,
+                    )
+
+                    if preview_from_txt2img:
+                        p.prompt = preview_prompt
+                        p.negative_prompt = preview_negative_prompt
+                        p.steps = preview_steps
+                        p.sampler_name = sd_samplers.samplers[preview_sampler_index].name
+                        p.cfg_scale = preview_cfg_scale
+                        p.seed = preview_seed
+                        p.width = preview_width
+                        p.height = preview_height
+                    else:
+                        p.prompt = batch.cond_text[0]
+                        p.steps = 20
+                        p.width = training_width
+                        p.height = training_height
+
+                    preview_text = p.prompt
+
+                    processed = processing.process_images(p)
+                    image = processed.images[0] if len(processed.images) > 0 else None
+
+                    if unload:
+                        shared.sd_model.first_stage_model.to(devices.cpu)
+
+                    if image is not None:
+                        shared.state.current_image = image
+                        last_saved_image, last_text_info = images.save_image(image, images_dir, "", p.seed, p.prompt, shared.opts.samples_format, processed.infotexts[0], p=p, forced_filename=forced_filename, save_to_dirs=False)
+                        last_saved_image += f", prompt: {preview_text}"
+
+                    if save_image_with_stored_embedding and os.path.exists(last_saved_file) and embedding_yet_to_be_embedded:
+
+                        last_saved_image_chunks = os.path.join(images_embeds_dir, f'{embedding_name}-{steps_done}.png')
+
+                        info = PngImagePlugin.PngInfo()
+                        data = torch.load(last_saved_file)
+                        info.add_text("sd-ti-embedding", embedding_to_b64(data))
+
+                        title = "<{}>".format(data.get('name', '???'))
+
+                        try:
+                            vectorSize = list(data['string_to_param'].values())[0].shape[0]
+                        except Exception as e:
+                            vectorSize = '?'
+
+                        checkpoint = sd_models.select_checkpoint()
+                        footer_left = checkpoint.model_name
+                        footer_mid = '[{}]'.format(checkpoint.hash)
+                        footer_right = '{}v {}s'.format(vectorSize, steps_done)
+
+                        captioned_image = caption_image_overlay(image, title, footer_left, footer_mid, footer_right)
+                        captioned_image = insert_image_data_embed(captioned_image, data)
+
+                        captioned_image.save(last_saved_image_chunks, "PNG", pnginfo=info)
+                        embedding_yet_to_be_embedded = False
+
+                    last_saved_image, last_text_info = images.save_image(image, images_dir, "", p.seed, p.prompt, shared.opts.samples_format, processed.infotexts[0], p=p, forced_filename=forced_filename, save_to_dirs=False)
+                    last_saved_image += f", prompt: {preview_text}"
+
+                shared.state.job_no = embedding.step
+
+                shared.state.textinfo = f"""
 <p>
-Loss: {losses.mean():.7f}<br/>
+Loss: {loss_step:.7f}<br/>
 Step: {embedding.step}<br/>
-Last prompt: {html.escape(entries[0].cond_text)}<br/>
+Last prompt: {html.escape(batch.cond_text[0])}<br/>
 Last saved embedding: {html.escape(last_saved_file)}<br/>
 Last saved image: {html.escape(last_saved_image)}<br/>
 </p>
 """
-
-    filename = os.path.join(shared.cmd_opts.embeddings_dir, f'{embedding_name}.pt')
-    save_embedding(embedding, checkpoint, embedding_name, filename, remove_cached_checksum=True)
-    shared.sd_model.first_stage_model.to(devices.device)
+        filename = os.path.join(shared.cmd_opts.embeddings_dir, f'{embedding_name}.pt')
+        save_embedding(embedding, checkpoint, embedding_name, filename, remove_cached_checksum=True)
+    except Exception:
+        print(traceback.format_exc(), file=sys.stderr)
+        pass
+    finally:
+        pbar.leave = False
+        pbar.close()
+        shared.sd_model.first_stage_model.to(devices.device)
 
     return embedding, filename
 
diff --git a/modules/ui.py b/modules/ui.py
index a5953fce..9d2a1cbf 100644
--- a/modules/ui.py
+++ b/modules/ui.py
@@ -1262,7 +1262,7 @@ def create_ui(wrap_gradio_gpu_call):
                         with gr.Column():
                             with gr.Row():
                                 interrupt_preprocessing = gr.Button("Interrupt")
-                                run_preprocess = gr.Button(value="Preprocess", variant='primary')
+                            run_preprocess = gr.Button(value="Preprocess", variant='primary')
 
                     process_split.change(
                         fn=lambda show: gr_show(show),
@@ -1289,6 +1289,7 @@ def create_ui(wrap_gradio_gpu_call):
                         hypernetwork_learn_rate = gr.Textbox(label='Hypernetwork Learning rate', placeholder="Hypernetwork Learning rate", value="0.00001")
 
                     batch_size = gr.Number(label='Batch size', value=1, precision=0)
+                    gradient_step = gr.Number(label='Gradient accumulation steps', value=1, precision=0)
                     dataset_directory = gr.Textbox(label='Dataset directory', placeholder="Path to directory with input images")
                     log_directory = gr.Textbox(label='Log directory', placeholder="Path to directory where to write outputs", value="textual_inversion")
                     template_file = gr.Textbox(label='Prompt template file', value=os.path.join(script_path, "textual_inversion_templates", "style_filewords.txt"))
@@ -1299,6 +1300,11 @@ def create_ui(wrap_gradio_gpu_call):
                     save_embedding_every = gr.Number(label='Save a copy of embedding to log directory every N steps, 0 to disable', value=500, precision=0)
                     save_image_with_stored_embedding = gr.Checkbox(label='Save images with embedding in PNG chunks', value=True)
                     preview_from_txt2img = gr.Checkbox(label='Read parameters (prompt, etc...) from txt2img tab when making previews', value=False)
+                    with gr.Row():
+                        shuffle_tags = gr.Checkbox(label="Shuffle tags by ',' when creating prompts.", value=False)
+                        tag_drop_out = gr.Slider(minimum=0, maximum=1, step=0.1, label="Drop out tags when creating prompts.", value=0)
+                    with gr.Row():
+                        latent_sampling_method = gr.Radio(label='Choose latent sampling method', value="once", choices=['once', 'deterministic', 'random'])
 
                     with gr.Row():
                         interrupt_training = gr.Button(value="Interrupt")
@@ -1387,11 +1393,15 @@ def create_ui(wrap_gradio_gpu_call):
                 train_embedding_name,
                 embedding_learn_rate,
                 batch_size,
+                gradient_step,
                 dataset_directory,
                 log_directory,
                 training_width,
                 training_height,
                 steps,
+                shuffle_tags,
+                tag_drop_out,
+                latent_sampling_method,
                 create_image_every,
                 save_embedding_every,
                 template_file,
@@ -1412,11 +1422,15 @@ def create_ui(wrap_gradio_gpu_call):
                 train_hypernetwork_name,
                 hypernetwork_learn_rate,
                 batch_size,
+                gradient_step,
                 dataset_directory,
                 log_directory,
                 training_width,
                 training_height,
                 steps,
+                shuffle_tags,
+                tag_drop_out,
+                latent_sampling_method,
                 create_image_every,
                 save_embedding_every,
                 template_file,
-- 
cgit v1.2.3


From a4a5735d0a80218e59f8a6e8401726f7209a6a8d Mon Sep 17 00:00:00 2001
From: flamelaw <flamelaw.com3d2@gmail.com>
Date: Sun, 20 Nov 2022 12:38:18 +0900
Subject: remove unnecessary comment

---
 modules/textual_inversion/dataset.py | 9 ---------
 1 file changed, 9 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index d594b49d..1dd53b85 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -103,15 +103,6 @@ class PersonalizedBase(Dataset):
             if include_cond and not (self.tag_drop_out != 0 or self.shuffle_tags):
                 with torch.autocast("cuda"):
                     entry.cond = cond_model([entry.cond_text]).to(devices.cpu).squeeze(0)
-            # elif not include_cond:
-            #     _, _, _, _, hijack_fixes, token_count = cond_model.process_text([entry.cond_text])
-            #     max_n = token_count // 75
-            #     index_list = [ [] for _ in range(max_n + 1) ]
-            #     for n, (z, _) in hijack_fixes[0]:
-            #         index_list[n].append(z)
-            #     with torch.autocast("cuda"):
-            #         entry.cond = cond_model([entry.cond_text]).to(devices.cpu).squeeze(0)
-            #     entry.emb_index = index_list
 
             self.dataset.append(entry)
             del torchdata
-- 
cgit v1.2.3


From 2d22d72cdaaf2b78b2986b841d478c11ac855dd2 Mon Sep 17 00:00:00 2001
From: flamelaw <flamelaw.com3d2@gmail.com>
Date: Sun, 20 Nov 2022 16:14:27 +0900
Subject: fix random sampling with pin_memory

---
 modules/textual_inversion/dataset.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index 1dd53b85..110c0e09 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -134,7 +134,7 @@ class PersonalizedBase(Dataset):
         if self.tag_drop_out != 0 or self.shuffle_tags:
             entry.cond_text = self.create_text(entry.filename_text)
         if self.latent_sampling_method == "random":
-            entry.latent_sample = shared.sd_model.get_first_stage_encoding(entry.latent_dist)
+            entry.latent_sample = shared.sd_model.get_first_stage_encoding(entry.latent_dist).to(devices.cpu)
         return entry
 
 class PersonalizedDataLoader(DataLoader):
-- 
cgit v1.2.3


From 5b57f61ba47f8b11d19a5b46e7fb5a52458abae5 Mon Sep 17 00:00:00 2001
From: flamelaw <flamelaw.com3d2@gmail.com>
Date: Mon, 21 Nov 2022 10:15:46 +0900
Subject: fix pin_memory with different latent sampling method

---
 modules/hypernetworks/hypernetwork.py          |  5 ++++-
 modules/textual_inversion/dataset.py           | 23 +++++++++++++++++++----
 modules/textual_inversion/textual_inversion.py |  7 +------
 3 files changed, 24 insertions(+), 11 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py
index 3d3301b0..0128419b 100644
--- a/modules/hypernetworks/hypernetwork.py
+++ b/modules/hypernetworks/hypernetwork.py
@@ -416,7 +416,10 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, gradient_step,
     pin_memory = shared.opts.pin_memory
 
     ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=hypernetwork_name, model=shared.sd_model, cond_model=shared.sd_model.cond_stage_model, device=devices.device, template_file=template_file, include_cond=True, batch_size=batch_size, gradient_step=gradient_step, shuffle_tags=shuffle_tags, tag_drop_out=tag_drop_out, latent_sampling_method=latent_sampling_method)
-    dl = modules.textual_inversion.dataset.PersonalizedDataLoader(ds, batch_size=ds.batch_size, pin_memory=pin_memory)
+    
+    latent_sampling_method = ds.latent_sampling_method
+
+    dl = modules.textual_inversion.dataset.PersonalizedDataLoader(ds, latent_sampling_method=latent_sampling_method, batch_size=ds.batch_size, pin_memory=pin_memory)
 
     if unload:
         shared.sd_model.cond_stage_model.to(devices.cpu)
diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index 110c0e09..f470324a 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -138,9 +138,12 @@ class PersonalizedBase(Dataset):
         return entry
 
 class PersonalizedDataLoader(DataLoader):
-    def __init__(self, *args, **kwargs):
-        super(PersonalizedDataLoader, self).__init__(shuffle=True, drop_last=True, *args, **kwargs)
-        self.collate_fn = collate_wrapper
+    def __init__(self, dataset, latent_sampling_method="once", batch_size=1, pin_memory=False):
+        super(PersonalizedDataLoader, self).__init__(dataset, shuffle=True, drop_last=True, batch_size=batch_size, pin_memory=pin_memory)
+        if latent_sampling_method == "random":
+            self.collate_fn = collate_wrapper_random
+        else:
+            self.collate_fn = collate_wrapper
         
 
 class BatchLoader:
@@ -148,10 +151,22 @@ class BatchLoader:
         self.cond_text = [entry.cond_text for entry in data]
         self.cond = [entry.cond for entry in data]
         self.latent_sample = torch.stack([entry.latent_sample for entry in data]).squeeze(1)
+        #self.emb_index = [entry.emb_index for entry in data]
+        #print(self.latent_sample.device)
 
     def pin_memory(self):
         self.latent_sample = self.latent_sample.pin_memory()
         return self
 
 def collate_wrapper(batch):
-    return BatchLoader(batch)
\ No newline at end of file
+    return BatchLoader(batch)
+
+class BatchLoaderRandom(BatchLoader):
+    def __init__(self, data):
+        super().__init__(data)
+
+    def pin_memory(self):
+        return self
+
+def collate_wrapper_random(batch):
+    return BatchLoaderRandom(batch)
\ No newline at end of file
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index 1d5e3a32..3036e48a 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -277,7 +277,7 @@ def train_embedding(embedding_name, learn_rate, batch_size, gradient_step, data_
 
     latent_sampling_method = ds.latent_sampling_method
 
-    dl = modules.textual_inversion.dataset.PersonalizedDataLoader(ds, batch_size=ds.batch_size, pin_memory=False)
+    dl = modules.textual_inversion.dataset.PersonalizedDataLoader(ds, latent_sampling_method=latent_sampling_method, batch_size=ds.batch_size, pin_memory=pin_memory)
 
     if unload:
         shared.sd_model.first_stage_model.to(devices.cpu)
@@ -333,11 +333,6 @@ def train_embedding(embedding_name, learn_rate, batch_size, gradient_step, data_
                 # go back until we reach gradient accumulation steps
                 if (j + 1) % gradient_step != 0:
                     continue
-                #print(f"grad:{embedding.vec.grad.detach().cpu().abs().mean().item():.7f}")
-                #scaler.unscale_(optimizer)
-                #print(f"grad:{embedding.vec.grad.detach().cpu().abs().mean().item():.7f}")
-                #torch.nn.utils.clip_grad_norm_(embedding.vec, max_norm=1.0)
-                #print(f"grad:{embedding.vec.grad.detach().cpu().abs().mean().item():.7f}")
                 scaler.step(optimizer)
                 scaler.update()
                 embedding.step += 1
-- 
cgit v1.2.3


From 4d5f1691dda971ec7b461dd880426300fd54ccee Mon Sep 17 00:00:00 2001
From: brkirch <brkirch@users.noreply.github.com>
Date: Mon, 28 Nov 2022 21:36:35 -0500
Subject: Use devices.autocast instead of torch.autocast

---
 modules/hypernetworks/hypernetwork.py          | 2 +-
 modules/interrogate.py                         | 3 +--
 modules/swinir_model.py                        | 6 +-----
 modules/textual_inversion/dataset.py           | 4 ++--
 modules/textual_inversion/textual_inversion.py | 2 +-
 5 files changed, 6 insertions(+), 11 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py
index 8466887f..eb5ae372 100644
--- a/modules/hypernetworks/hypernetwork.py
+++ b/modules/hypernetworks/hypernetwork.py
@@ -495,7 +495,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, gradient_step,
                 if shared.state.interrupted:
                     break
 
-                with torch.autocast("cuda"):
+                with devices.autocast():
                     x = batch.latent_sample.to(devices.device, non_blocking=pin_memory)
                     if tag_drop_out != 0 or shuffle_tags:
                         shared.sd_model.cond_stage_model.to(devices.device)
diff --git a/modules/interrogate.py b/modules/interrogate.py
index 9769aa34..40c6b082 100644
--- a/modules/interrogate.py
+++ b/modules/interrogate.py
@@ -148,8 +148,7 @@ class InterrogateModels:
 
             clip_image = self.clip_preprocess(pil_image).unsqueeze(0).type(self.dtype).to(devices.device_interrogate)
 
-            precision_scope = torch.autocast if shared.cmd_opts.precision == "autocast" else contextlib.nullcontext
-            with torch.no_grad(), precision_scope("cuda"):
+            with torch.no_grad(), devices.autocast():
                 image_features = self.clip_model.encode_image(clip_image).type(self.dtype)
 
                 image_features /= image_features.norm(dim=-1, keepdim=True)
diff --git a/modules/swinir_model.py b/modules/swinir_model.py
index facd262d..483eabd4 100644
--- a/modules/swinir_model.py
+++ b/modules/swinir_model.py
@@ -13,10 +13,6 @@ from modules.swinir_model_arch import SwinIR as net
 from modules.swinir_model_arch_v2 import Swin2SR as net2
 from modules.upscaler import Upscaler, UpscalerData
 
-precision_scope = (
-    torch.autocast if cmd_opts.precision == "autocast" else contextlib.nullcontext
-)
-
 
 class UpscalerSwinIR(Upscaler):
     def __init__(self, dirname):
@@ -112,7 +108,7 @@ def upscale(
     img = np.moveaxis(img, 2, 0) / 255
     img = torch.from_numpy(img).float()
     img = img.unsqueeze(0).to(devices.device_swinir)
-    with torch.no_grad(), precision_scope("cuda"):
+    with torch.no_grad(), devices.autocast():
         _, _, h_old, w_old = img.size()
         h_pad = (h_old // window_size + 1) * window_size - h_old
         w_pad = (w_old // window_size + 1) * window_size - w_old
diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index e5725f33..2dc64c3c 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -82,7 +82,7 @@ class PersonalizedBase(Dataset):
             torchdata = torch.from_numpy(npimage).permute(2, 0, 1).to(device=device, dtype=torch.float32)
             latent_sample = None
 
-            with torch.autocast("cuda"):
+            with devices.autocast():
                 latent_dist = model.encode_first_stage(torchdata.unsqueeze(dim=0))
 
             if latent_sampling_method == "once" or (latent_sampling_method == "deterministic" and not isinstance(latent_dist, DiagonalGaussianDistribution)):
@@ -101,7 +101,7 @@ class PersonalizedBase(Dataset):
                 entry.cond_text = self.create_text(filename_text)
 
             if include_cond and not (self.tag_drop_out != 0 or self.shuffle_tags):
-                with torch.autocast("cuda"):
+                with devices.autocast():
                     entry.cond = cond_model([entry.cond_text]).to(devices.cpu).squeeze(0)
 
             self.dataset.append(entry)
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index 4eb75cb5..daf8d1b8 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -316,7 +316,7 @@ def train_embedding(embedding_name, learn_rate, batch_size, gradient_step, data_
                 if shared.state.interrupted:
                     break
 
-                with torch.autocast("cuda"):
+                with devices.autocast():
                     # c = stack_conds(batch.cond).to(devices.device)
                     # mask = torch.tensor(batch.emb_index).to(devices.device, non_blocking=pin_memory)
                     # print(mask)
-- 
cgit v1.2.3


From c0355caefe3d82e304e6d832699d581fc8f9fbf9 Mon Sep 17 00:00:00 2001
From: Jim Hays <jim@emporatitle.com>
Date: Wed, 14 Dec 2022 21:01:32 -0500
Subject: Fix various typos

---
 README.md                                      |  4 ++--
 javascript/contextMenus.js                     | 24 ++++++++++++------------
 javascript/progressbar.js                      | 12 ++++++------
 javascript/ui.js                               |  2 +-
 modules/api/api.py                             | 18 +++++++++---------
 modules/api/models.py                          |  2 +-
 modules/images.py                              |  4 ++--
 modules/processing.py                          | 14 +++++++-------
 modules/safe.py                                |  4 ++--
 modules/scripts.py                             |  4 ++--
 modules/sd_hijack_inpainting.py                |  6 +++---
 modules/sd_hijack_unet.py                      |  2 +-
 modules/textual_inversion/dataset.py           | 10 +++++-----
 modules/textual_inversion/textual_inversion.py | 16 ++++++++--------
 scripts/prompt_matrix.py                       | 10 +++++-----
 webui.py                                       |  4 ++--
 16 files changed, 68 insertions(+), 68 deletions(-)

(limited to 'modules/textual_inversion/dataset.py')

diff --git a/README.md b/README.md
index 55990581..556000fb 100644
--- a/README.md
+++ b/README.md
@@ -82,8 +82,8 @@ Check the [custom scripts](https://github.com/AUTOMATIC1111/stable-diffusion-web
 - Use VAEs
 - Estimated completion time in progress bar
 - API
-- Support for dedicated [inpainting model](https://github.com/runwayml/stable-diffusion#inpainting-with-stable-diffusion) by RunwayML. 
-- via extension: [Aesthetic Gradients](https://github.com/AUTOMATIC1111/stable-diffusion-webui-aesthetic-gradients), a way to generate images with a specific aesthetic by using clip images embds (implementation of [https://github.com/vicgalle/stable-diffusion-aesthetic-gradients](https://github.com/vicgalle/stable-diffusion-aesthetic-gradients))
+- Support for dedicated [inpainting model](https://github.com/runwayml/stable-diffusion#inpainting-with-stable-diffusion) by RunwayML.
+- via extension: [Aesthetic Gradients](https://github.com/AUTOMATIC1111/stable-diffusion-webui-aesthetic-gradients), a way to generate images with a specific aesthetic by using clip images embeds (implementation of [https://github.com/vicgalle/stable-diffusion-aesthetic-gradients](https://github.com/vicgalle/stable-diffusion-aesthetic-gradients))
 - [Stable Diffusion 2.0](https://github.com/Stability-AI/stablediffusion) support - see [wiki](https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#stable-diffusion-20) for instructions
 
 ## Installation and Running
diff --git a/javascript/contextMenus.js b/javascript/contextMenus.js
index fe67c42e..11bcce1b 100644
--- a/javascript/contextMenus.js
+++ b/javascript/contextMenus.js
@@ -9,7 +9,7 @@ contextMenuInit = function(){
 
   function showContextMenu(event,element,menuEntries){
     let posx = event.clientX + document.body.scrollLeft + document.documentElement.scrollLeft;
-    let posy = event.clientY + document.body.scrollTop + document.documentElement.scrollTop; 
+    let posy = event.clientY + document.body.scrollTop + document.documentElement.scrollTop;
 
     let oldMenu = gradioApp().querySelector('#context-menu')
     if(oldMenu){
@@ -61,15 +61,15 @@ contextMenuInit = function(){
 
   }
 
-  function appendContextMenuOption(targetEmementSelector,entryName,entryFunction){
-    
-    currentItems = menuSpecs.get(targetEmementSelector)
-    
+  function appendContextMenuOption(targetElementSelector,entryName,entryFunction){
+
+    currentItems = menuSpecs.get(targetElementSelector)
+
     if(!currentItems){
       currentItems = []
-      menuSpecs.set(targetEmementSelector,currentItems);
+      menuSpecs.set(targetElementSelector,currentItems);
     }
-    let newItem = {'id':targetEmementSelector+'_'+uid(), 
+    let newItem = {'id':targetElementSelector+'_'+uid(),
                    'name':entryName,
                    'func':entryFunction,
                    'isNew':true}
@@ -97,7 +97,7 @@ contextMenuInit = function(){
       if(source.id && source.id.indexOf('check_progress')>-1){
         return
       }
-      
+
       let oldMenu = gradioApp().querySelector('#context-menu')
       if(oldMenu){
         oldMenu.remove()
@@ -117,7 +117,7 @@ contextMenuInit = function(){
       })
     });
     eventListenerApplied=true
-  
+
   }
 
   return [appendContextMenuOption, removeContextMenuOption, addContextMenuEventListener]
@@ -152,8 +152,8 @@ addContextMenuEventListener = initResponse[2];
     generateOnRepeat('#img2img_generate','#img2img_interrupt');
   })
 
-  let cancelGenerateForever = function(){ 
-    clearInterval(window.generateOnRepeatInterval) 
+  let cancelGenerateForever = function(){
+    clearInterval(window.generateOnRepeatInterval)
   }
 
   appendContextMenuOption('#txt2img_interrupt','Cancel generate forever',cancelGenerateForever)
@@ -162,7 +162,7 @@ addContextMenuEventListener = initResponse[2];
   appendContextMenuOption('#img2img_generate', 'Cancel generate forever',cancelGenerateForever)
 
   appendContextMenuOption('#roll','Roll three',
-    function(){ 
+    function(){
       let rollbutton = get_uiCurrentTabContent().querySelector('#roll');
       setTimeout(function(){rollbutton.click()},100)
       setTimeout(function(){rollbutton.click()},200)
diff --git a/javascript/progressbar.js b/javascript/progressbar.js
index d58737c4..d6323ed9 100644
--- a/javascript/progressbar.js
+++ b/javascript/progressbar.js
@@ -3,7 +3,7 @@ global_progressbars = {}
 galleries = {}
 galleryObservers = {}
 
-// this tracks laumnches of window.setTimeout for progressbar to prevent starting a new timeout when the previous is still running
+// this tracks launches of window.setTimeout for progressbar to prevent starting a new timeout when the previous is still running
 timeoutIds = {}
 
 function check_progressbar(id_part, id_progressbar, id_progressbar_span, id_skip, id_interrupt, id_preview, id_gallery){
@@ -20,21 +20,21 @@ function check_progressbar(id_part, id_progressbar, id_progressbar_span, id_skip
 
     var skip = id_skip ? gradioApp().getElementById(id_skip) : null
     var interrupt = gradioApp().getElementById(id_interrupt)
-    
+
     if(opts.show_progress_in_title && progressbar && progressbar.offsetParent){
         if(progressbar.innerText){
             let newtitle = '[' + progressbar.innerText.trim() + '] Stable Diffusion';
             if(document.title != newtitle){
-                document.title =  newtitle;          
+                document.title =  newtitle;
             }
         }else{
             let newtitle = 'Stable Diffusion'
             if(document.title != newtitle){
-                document.title =  newtitle;          
+                document.title =  newtitle;
             }
         }
     }
-    
+
 	if(progressbar!= null && progressbar != global_progressbars[id_progressbar]){
 	    global_progressbars[id_progressbar] = progressbar
 
@@ -63,7 +63,7 @@ function check_progressbar(id_part, id_progressbar, id_progressbar_span, id_skip
                         skip.style.display = "none"
                     }
                     interrupt.style.display = "none"
-			
+
                     //disconnect observer once generation finished, so user can close selected image if they want
                     if (galleryObservers[id_gallery]) {
                         galleryObservers[id_gallery].disconnect();
diff --git a/javascript/ui.js b/javascript/ui.js
index 2cb280e5..587dd782 100644
--- a/javascript/ui.js
+++ b/javascript/ui.js
@@ -100,7 +100,7 @@ function create_submit_args(args){
 
     // As it is currently, txt2img and img2img send back the previous output args (txt2img_gallery, generation_info, html_info) whenever you generate a new image.
     // This can lead to uploading a huge gallery of previously generated images, which leads to an unnecessary delay between submitting and beginning to generate.
-    // I don't know why gradio is seding outputs along with inputs, but we can prevent sending the image gallery here, which seems to be an issue for some.
+    // I don't know why gradio is sending outputs along with inputs, but we can prevent sending the image gallery here, which seems to be an issue for some.
     // If gradio at some point stops sending outputs, this may break something
     if(Array.isArray(res[res.length - 3])){
         res[res.length - 3] = null
diff --git a/modules/api/api.py b/modules/api/api.py
index 89935a70..33845045 100644
--- a/modules/api/api.py
+++ b/modules/api/api.py
@@ -67,10 +67,10 @@ def encode_pil_to_base64(image):
 class Api:
     def __init__(self, app: FastAPI, queue_lock: Lock):
         if shared.cmd_opts.api_auth:
-            self.credenticals = dict()
+            self.credentials = dict()
             for auth in shared.cmd_opts.api_auth.split(","):
                 user, password = auth.split(":")
-                self.credenticals[user] = password
+                self.credentials[user] = password
 
         self.router = APIRouter()
         self.app = app
@@ -93,7 +93,7 @@ class Api:
         self.add_api_route("/sdapi/v1/hypernetworks", self.get_hypernetworks, methods=["GET"], response_model=List[HypernetworkItem])
         self.add_api_route("/sdapi/v1/face-restorers", self.get_face_restorers, methods=["GET"], response_model=List[FaceRestorerItem])
         self.add_api_route("/sdapi/v1/realesrgan-models", self.get_realesrgan_models, methods=["GET"], response_model=List[RealesrganItem])
-        self.add_api_route("/sdapi/v1/prompt-styles", self.get_promp_styles, methods=["GET"], response_model=List[PromptStyleItem])
+        self.add_api_route("/sdapi/v1/prompt-styles", self.get_prompt_styles, methods=["GET"], response_model=List[PromptStyleItem])
         self.add_api_route("/sdapi/v1/artist-categories", self.get_artists_categories, methods=["GET"], response_model=List[str])
         self.add_api_route("/sdapi/v1/artists", self.get_artists, methods=["GET"], response_model=List[ArtistItem])
 
@@ -102,9 +102,9 @@ class Api:
             return self.app.add_api_route(path, endpoint, dependencies=[Depends(self.auth)], **kwargs)
         return self.app.add_api_route(path, endpoint, **kwargs)
 
-    def auth(self, credenticals: HTTPBasicCredentials = Depends(HTTPBasic())):
-        if credenticals.username in self.credenticals:
-            if compare_digest(credenticals.password, self.credenticals[credenticals.username]):
+    def auth(self, credentials: HTTPBasicCredentials = Depends(HTTPBasic())):
+        if credentials.username in self.credentials:
+            if compare_digest(credentials.password, self.credentials[credentials.username]):
                 return True
 
         raise HTTPException(status_code=401, detail="Incorrect username or password", headers={"WWW-Authenticate": "Basic"})
@@ -239,7 +239,7 @@ class Api:
     def interrogateapi(self, interrogatereq: InterrogateRequest):
         image_b64 = interrogatereq.image
         if image_b64 is None:
-            raise HTTPException(status_code=404, detail="Image not found") 
+            raise HTTPException(status_code=404, detail="Image not found")
 
         img = decode_base64_to_image(image_b64)
         img = img.convert('RGB')
@@ -252,7 +252,7 @@ class Api:
                 processed = deepbooru.model.tag(img)
             else:
                 raise HTTPException(status_code=404, detail="Model not found")
-        
+
         return InterrogateResponse(caption=processed)
 
     def interruptapi(self):
@@ -308,7 +308,7 @@ class Api:
     def get_realesrgan_models(self):
         return [{"name":x.name,"path":x.data_path, "scale":x.scale} for x in get_realesrgan_models(None)]
 
-    def get_promp_styles(self):
+    def get_prompt_styles(self):
         styleList = []
         for k in shared.prompt_styles.styles:
             style = shared.prompt_styles.styles[k]
diff --git a/modules/api/models.py b/modules/api/models.py
index f77951fc..a22bc6b3 100644
--- a/modules/api/models.py
+++ b/modules/api/models.py
@@ -128,7 +128,7 @@ class ExtrasBaseRequest(BaseModel):
     upscaling_resize: float = Field(default=2, title="Upscaling Factor", ge=1, le=4, description="By how much to upscale the image, only used when resize_mode=0.")
     upscaling_resize_w: int = Field(default=512, title="Target Width", ge=1, description="Target width for the upscaler to hit. Only used when resize_mode=1.")
     upscaling_resize_h: int = Field(default=512, title="Target Height", ge=1, description="Target height for the upscaler to hit. Only used when resize_mode=1.")
-    upscaling_crop: bool = Field(default=True, title="Crop to fit", description="Should the upscaler crop the image to fit in the choosen size?")
+    upscaling_crop: bool = Field(default=True, title="Crop to fit", description="Should the upscaler crop the image to fit in the chosen size?")
     upscaler_1: str = Field(default="None", title="Main upscaler", description=f"The name of the main upscaler to use, it has to be one of this list: {' , '.join([x.name for x in sd_upscalers])}")
     upscaler_2: str = Field(default="None", title="Secondary upscaler", description=f"The name of the secondary upscaler to use, it has to be one of this list: {' , '.join([x.name for x in sd_upscalers])}")
     extras_upscaler_2_visibility: float = Field(default=0, title="Secondary upscaler visibility", ge=0, le=1, allow_inf_nan=False, description="Sets the visibility of secondary upscaler, values should be between 0 and 1.")
diff --git a/modules/images.py b/modules/images.py
index 8146f580..93a14289 100644
--- a/modules/images.py
+++ b/modules/images.py
@@ -429,7 +429,7 @@ def save_image(image, path, basename, seed=None, prompt=None, extension='png', i
             The directory to save the image. Note, the option `save_to_dirs` will make the image to be saved into a sub directory.
         basename (`str`):
             The base filename which will be applied to `filename pattern`.
-        seed, prompt, short_filename, 
+        seed, prompt, short_filename,
         extension (`str`):
             Image file extension, default is `png`.
         pngsectionname (`str`):
@@ -590,7 +590,7 @@ def read_info_from_image(image):
 Negative prompt: {json_info["uc"]}
 Steps: {json_info["steps"]}, Sampler: {sampler}, CFG scale: {json_info["scale"]}, Seed: {json_info["seed"]}, Size: {image.width}x{image.height}, Clip skip: 2, ENSD: 31337"""
         except Exception:
-            print(f"Error parsing NovelAI iamge generation parameters:", file=sys.stderr)
+            print(f"Error parsing NovelAI image generation parameters:", file=sys.stderr)
             print(traceback.format_exc(), file=sys.stderr)
 
     return geninfo, items
diff --git a/modules/processing.py b/modules/processing.py
index 24c537d1..fe7f4faf 100644
--- a/modules/processing.py
+++ b/modules/processing.py
@@ -147,11 +147,11 @@ class StableDiffusionProcessing():
 
         # The "masked-image" in this case will just be all zeros since the entire image is masked.
         image_conditioning = torch.zeros(x.shape[0], 3, height, width, device=x.device)
-        image_conditioning = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(image_conditioning)) 
+        image_conditioning = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(image_conditioning))
 
         # Add the fake full 1s mask to the first dimension.
         image_conditioning = torch.nn.functional.pad(image_conditioning, (0, 0, 0, 0, 1, 0), value=1.0)
-        image_conditioning = image_conditioning.to(x.dtype)            
+        image_conditioning = image_conditioning.to(x.dtype)
 
         return image_conditioning
 
@@ -199,7 +199,7 @@ class StableDiffusionProcessing():
             source_image * (1.0 - conditioning_mask),
             getattr(self, "inpainting_mask_weight", shared.opts.inpainting_mask_weight)
         )
-        
+
         # Encode the new masked image using first stage of network.
         conditioning_image = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(conditioning_image))
 
@@ -537,7 +537,7 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed:
         for n in range(p.n_iter):
             if state.skipped:
                 state.skipped = False
-            
+
             if state.interrupted:
                 break
 
@@ -612,7 +612,7 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed:
                     image.info["parameters"] = text
                 output_images.append(image)
 
-            del x_samples_ddim 
+            del x_samples_ddim
 
             devices.torch_gc()
 
@@ -704,7 +704,7 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing):
 
         samples = samples[:, :, self.truncate_y//2:samples.shape[2]-self.truncate_y//2, self.truncate_x//2:samples.shape[3]-self.truncate_x//2]
 
-        """saves image before applying hires fix, if enabled in options; takes as an arguyment either an image or batch with latent space images"""
+        """saves image before applying hires fix, if enabled in options; takes as an argument either an image or batch with latent space images"""
         def save_intermediate(image, index):
             if not opts.save or self.do_not_save_samples or not opts.save_images_before_highres_fix:
                 return
@@ -720,7 +720,7 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing):
 
             samples = torch.nn.functional.interpolate(samples, size=(self.height // opt_f, self.width // opt_f), mode="bilinear")
 
-            # Avoid making the inpainting conditioning unless necessary as 
+            # Avoid making the inpainting conditioning unless necessary as
             # this does need some extra compute to decode / encode the image again.
             if getattr(self, "inpainting_mask_weight", shared.opts.inpainting_mask_weight) < 1.0:
                 image_conditioning = self.img2img_image_conditioning(decode_first_stage(self.sd_model, samples), samples)
diff --git a/modules/safe.py b/modules/safe.py
index 10460ad0..20e9d2fa 100644
--- a/modules/safe.py
+++ b/modules/safe.py
@@ -80,7 +80,7 @@ def check_pt(filename, extra_handler):
         # new pytorch format is a zip file
         with zipfile.ZipFile(filename) as z:
             check_zip_filenames(filename, z.namelist())
-            
+
             # find filename of data.pkl in zip file: '<directory name>/data.pkl'
             data_pkl_filenames = [f for f in z.namelist() if data_pkl_re.match(f)]
             if len(data_pkl_filenames) == 0:
@@ -108,7 +108,7 @@ def load(filename, *args, **kwargs):
 
 def load_with_extra(filename, extra_handler=None, *args, **kwargs):
     """
-    this functon is intended to be used by extensions that want to load models with
+    this function is intended to be used by extensions that want to load models with
     some extra classes in them that the usual unpickler would find suspicious.
 
     Use the extra_handler argument to specify a function that takes module and field name as text,
diff --git a/modules/scripts.py b/modules/scripts.py
index 23ca195d..722f8685 100644
--- a/modules/scripts.py
+++ b/modules/scripts.py
@@ -36,7 +36,7 @@ class Script:
     def ui(self, is_img2img):
         """this function should create gradio UI elements. See https://gradio.app/docs/#components
         The return value should be an array of all components that are used in processing.
-        Values of those returned componenbts will be passed to run() and process() functions.
+        Values of those returned components will be passed to run() and process() functions.
         """
 
         pass
@@ -47,7 +47,7 @@ class Script:
 
         This function should return:
          - False if the script should not be shown in UI at all
-         - True if the script should be shown in UI if it's scelected in the scripts drowpdown
+         - True if the script should be shown in UI if it's selected in the scripts dropdown
          - script.AlwaysVisible if the script should be shown in UI at all times
          """
 
diff --git a/modules/sd_hijack_inpainting.py b/modules/sd_hijack_inpainting.py
index 938f9a58..d72f83fd 100644
--- a/modules/sd_hijack_inpainting.py
+++ b/modules/sd_hijack_inpainting.py
@@ -209,7 +209,7 @@ def p_sample_plms(self, x, c, t, index, repeat_noise=False, use_original_steps=F
         else:
             x_in = torch.cat([x] * 2)
             t_in = torch.cat([t] * 2)
-            
+
             if isinstance(c, dict):
                 assert isinstance(unconditional_conditioning, dict)
                 c_in = dict()
@@ -278,7 +278,7 @@ def p_sample_plms(self, x, c, t, index, repeat_noise=False, use_original_steps=F
     x_prev, pred_x0 = get_x_prev_and_pred_x0(e_t_prime, index)
 
     return x_prev, pred_x0, e_t
-    
+
 # =================================================================================================
 # Monkey patch LatentInpaintDiffusion to load the checkpoint with a proper config.
 # Adapted from:
@@ -326,7 +326,7 @@ def do_inpainting_hijack():
     # most of this stuff seems to no longer be needed because it is already included into SD2.0
     # LatentInpaintDiffusion remains because SD2.0's LatentInpaintDiffusion can't be loaded without specifying a checkpoint
     # p_sample_plms is needed because PLMS can't work with dicts as conditionings
-    # this file should be cleaned up later if weverything tuens out to work fine
+    # this file should be cleaned up later if everything turns out to work fine
 
     # ldm.models.diffusion.ddpm.get_unconditional_conditioning = get_unconditional_conditioning
     ldm.models.diffusion.ddpm.LatentInpaintDiffusion = LatentInpaintDiffusion
diff --git a/modules/sd_hijack_unet.py b/modules/sd_hijack_unet.py
index 1b9d7757..18daf8c1 100644
--- a/modules/sd_hijack_unet.py
+++ b/modules/sd_hijack_unet.py
@@ -4,7 +4,7 @@ import torch
 class TorchHijackForUnet:
     """
     This is torch, but with cat that resizes tensors to appropriate dimensions if they do not match;
-    this makes it possible to create pictures with dimensions that are muliples of 8 rather than 64
+    this makes it possible to create pictures with dimensions that are multiples of 8 rather than 64
     """
 
     def __getattr__(self, item):
diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py
index 2dc64c3c..88d68c76 100644
--- a/modules/textual_inversion/dataset.py
+++ b/modules/textual_inversion/dataset.py
@@ -28,9 +28,9 @@ class DatasetEntry:
 
 
 class PersonalizedBase(Dataset):
-    def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", model=None, cond_model=None, device=None, template_file=None, include_cond=False, batch_size=1, gradient_step=1, shuffle_tags=False, tag_drop_out=0, latent_sampling_method='once'):        
+    def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", model=None, cond_model=None, device=None, template_file=None, include_cond=False, batch_size=1, gradient_step=1, shuffle_tags=False, tag_drop_out=0, latent_sampling_method='once'):
         re_word = re.compile(shared.opts.dataset_filename_word_regex) if len(shared.opts.dataset_filename_word_regex) > 0 else None
-        
+
         self.placeholder_token = placeholder_token
 
         self.width = width
@@ -50,14 +50,14 @@ class PersonalizedBase(Dataset):
 
         self.image_paths = [os.path.join(data_root, file_path) for file_path in os.listdir(data_root)]
 
-        
+
         self.shuffle_tags = shuffle_tags
         self.tag_drop_out = tag_drop_out
 
         print("Preparing dataset...")
         for path in tqdm.tqdm(self.image_paths):
             if shared.state.interrupted:
-                raise Exception("inturrupted")
+                raise Exception("interrupted")
             try:
                 image = Image.open(path).convert('RGB').resize((self.width, self.height), PIL.Image.BICUBIC)
             except Exception:
@@ -144,7 +144,7 @@ class PersonalizedDataLoader(DataLoader):
             self.collate_fn = collate_wrapper_random
         else:
             self.collate_fn = collate_wrapper
-        
+
 
 class BatchLoader:
     def __init__(self, data):
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index e28c357a..daf3997b 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -133,7 +133,7 @@ class EmbeddingDatabase:
 
                 process_file(fullfn, fn)
             except Exception:
-                print(f"Error loading emedding {fn}:", file=sys.stderr)
+                print(f"Error loading embedding {fn}:", file=sys.stderr)
                 print(traceback.format_exc(), file=sys.stderr)
                 continue
 
@@ -194,7 +194,7 @@ def write_loss(log_directory, filename, step, epoch_len, values):
             csv_writer.writeheader()
 
         epoch = (step - 1) // epoch_len
-        epoch_step = (step - 1) % epoch_len 
+        epoch_step = (step - 1) % epoch_len
 
         csv_writer.writerow({
             "step": step,
@@ -270,9 +270,9 @@ def train_embedding(embedding_name, learn_rate, batch_size, gradient_step, data_
    # dataset loading may take a while, so input validations and early returns should be done before this
     shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..."
     old_parallel_processing_allowed = shared.parallel_processing_allowed
-    
+
     pin_memory = shared.opts.pin_memory
-    
+
     ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=embedding_name, model=shared.sd_model, cond_model=shared.sd_model.cond_stage_model, device=devices.device, template_file=template_file, batch_size=batch_size, gradient_step=gradient_step, shuffle_tags=shuffle_tags, tag_drop_out=tag_drop_out, latent_sampling_method=latent_sampling_method)
 
     latent_sampling_method = ds.latent_sampling_method
@@ -295,12 +295,12 @@ def train_embedding(embedding_name, learn_rate, batch_size, gradient_step, data_
     loss_step = 0
     _loss_step = 0 #internal
 
-    
+
     last_saved_file = "<none>"
     last_saved_image = "<none>"
     forced_filename = "<none>"
     embedding_yet_to_be_embedded = False
-    
+
     pbar = tqdm.tqdm(total=steps - initial_step)
     try:
         for i in range((steps-initial_step) * gradient_step):
@@ -327,10 +327,10 @@ def train_embedding(embedding_name, learn_rate, batch_size, gradient_step, data_
                     c = shared.sd_model.cond_stage_model(batch.cond_text)
                     loss = shared.sd_model(x, c)[0] / gradient_step
                     del x
-                    
+
                     _loss_step += loss.item()
                 scaler.scale(loss).backward()
-                
+
                 # go back until we reach gradient accumulation steps
                 if (j + 1) % gradient_step != 0:
                     continue
diff --git a/scripts/prompt_matrix.py b/scripts/prompt_matrix.py
index c53ca28c..4c79eaef 100644
--- a/scripts/prompt_matrix.py
+++ b/scripts/prompt_matrix.py
@@ -18,7 +18,7 @@ def draw_xy_grid(xs, ys, x_label, y_label, cell):
     ver_texts = [[images.GridAnnotation(y_label(y))] for y in ys]
     hor_texts = [[images.GridAnnotation(x_label(x))] for x in xs]
 
-    first_pocessed = None
+    first_processed = None
 
     state.job_count = len(xs) * len(ys)
 
@@ -27,17 +27,17 @@ def draw_xy_grid(xs, ys, x_label, y_label, cell):
             state.job = f"{ix + iy * len(xs) + 1} out of {len(xs) * len(ys)}"
 
             processed = cell(x, y)
-            if first_pocessed is None:
-                first_pocessed = processed
+            if first_processed is None:
+                first_processed = processed
 
             res.append(processed.images[0])
 
     grid = images.image_grid(res, rows=len(ys))
     grid = images.draw_grid_annotations(grid, res[0].width, res[0].height, hor_texts, ver_texts)
 
-    first_pocessed.images = [grid]
+    first_processed.images = [grid]
 
-    return first_pocessed
+    return first_processed
 
 
 class Script(scripts.Script):
diff --git a/webui.py b/webui.py
index c2d0c6be..4b32e77d 100644
--- a/webui.py
+++ b/webui.py
@@ -153,8 +153,8 @@ def webui():
 
         # gradio uses a very open CORS policy via app.user_middleware, which makes it possible for
         # an attacker to trick the user into opening a malicious HTML page, which makes a request to the
-        # running web ui and do whatever the attcker wants, including installing an extension and
-        # runnnig its code. We disable this here. Suggested by RyotaK.
+        # running web ui and do whatever the attacker wants, including installing an extension and
+        # running its code. We disable this here. Suggested by RyotaK.
         app.user_middleware = [x for x in app.user_middleware if x.cls.__name__ != 'CORSMiddleware']
 
         setup_cors(app)
-- 
cgit v1.2.3