hoelee/stable-diffusion-webui
1
0
Fork 0
mirror of https://github.com/AUTOMATIC1111/stable-diffusion-webui.git synced 2025-08-08 05:12:35 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Textual Inversion: Added custom training image size and number of repeats per input image in a single epoch

Browse Source
This commit is contained in:
alg-wiki
2022-10-10 17:07:46 +09:00
committed by AUTOMATIC1111
parent 8f1efdc130
commit ea00c1624b
4 changed files with 24 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
modules/textual_inversion/dataset.py
View File

@@ -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 = []

4
modules/textual_inversion/preprocess.py
View File

@@ -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)

15
modules/textual_inversion/textual_inversion.py
View File

@@ -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,
)