File size: 10,331 Bytes
3dabe4a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 |
NAME = 'XL Vec'
from torch import Tensor, FloatTensor, nn
import gradio as gr
from modules.processing import StableDiffusionProcessing
from modules import scripts
from scripts.sdhook import SDHook
from scripts.xl_clip import CLIP_SDXL, get_pooled
from scripts.xl_vec_xyz import init_xyz
def hook_input(
args: 'Hook',
mod: nn.Module,
inputs: tuple[dict[str,Tensor]]
):
if not args.enabled:
return
assert isinstance(mod, CLIP_SDXL)
input = inputs[0]
overwritten_keys = set()
def create(v: list[float], src: FloatTensor):
return FloatTensor(v).to(dtype=src.dtype, device=src.device)
def put(name: str, v: list[float]):
if name in input:
src = input[name]
input[name] = create(v, src).reshape(src.shape)
overwritten_keys.add(name)
old = {k: v for k, v in input.items()}
put('original_size_as_tuple', [args.original_height, args.original_width])
put('crop_coords_top_left', [args.crop_top, args.crop_left])
put('target_size_as_tuple', [args.target_height, args.target_width])
if input['aesthetic_score'].item() == 6.0:
# positive prompt
put('aesthetic_score', [args.aesthetic_score])
else:
# negative prompt
put('aesthetic_score', [args.negative_aesthetic_score])
new = {k: v for k, v in input.items()}
for k in overwritten_keys:
print(f"{k}: {old[k].tolist()} -> {new[k].tolist()}")
return inputs
def hook_output(
args: 'Hook',
mod: nn.Module,
inputs: tuple[dict[str,Tensor]],
output: dict,
):
if not args.enabled:
return
if inputs[0]['aesthetic_score'].item() == 6.0:
# positive prompt
prompt = args.extra_prompt
index = args.token_index
multiplier = args.eot_multiplier
else:
# negative prompt
prompt = args.extra_negative_prompt
index = args.negative_token_index
multiplier = args.negative_eot_multiplier
if prompt is None or len(prompt) == 0:
if index == -1 and multiplier == 1.0:
# default
return
# use original prompt
prompt = inputs[0]['txt'][0]
assert isinstance(mod, CLIP_SDXL)
try:
args.enabled = False
pooled, at = get_pooled(mod, prompt, index=index) # (1,1280)
assert pooled.shape == (1, 1280), f'pooled.shape={pooled.shape}'
finally:
args.enabled = True
output['vector'][:, 0:1280] = pooled[:] * multiplier
print(f"vector[:, 0:1280]: {inputs[0]['txt']} -> {[prompt]} @ {at} [M={multiplier:.3f}]")
return output
class Hook(SDHook):
def __init__(
self,
enabled: bool,
p: StableDiffusionProcessing,
crop_left: float,
crop_top: float,
original_width: float,
original_height: float,
target_width: float,
target_height: float,
aesthetic_score: float,
negative_aesthetic_score: float,
extra_prompt: str|None,
extra_negative_prompt: str|None,
token_index: int|float,
negative_token_index: int|float,
eot_multiplier: float,
negative_eot_multiplier: float,
with_hr: bool,
):
super().__init__(enabled)
self.p = p
self.crop_left = float(crop_left)
self.crop_top = float(crop_top)
self.original_width = float(original_width)
self.original_height = float(original_height)
self.target_width = float(target_width)
self.target_height = float(target_height)
self.aesthetic_score = float(aesthetic_score)
self.negative_aesthetic_score = float(negative_aesthetic_score)
self.extra_prompt = extra_prompt
self.extra_negative_prompt = extra_negative_prompt
self.token_index = int(token_index)
self.negative_token_index = int(negative_token_index)
self.eot_multiplier = float(eot_multiplier)
self.negative_eot_multiplier = float(negative_eot_multiplier)
self.with_hr = bool(with_hr)
def hook_clip(self, p: StableDiffusionProcessing, clip: nn.Module):
if not hasattr(p.sd_model, 'is_sdxl') or not p.sd_model.is_sdxl:
return
def inp(*args, **kwargs):
return hook_input(self, *args, **kwargs)
def outp(*args, **kwargs):
return hook_output(self, *args, **kwargs)
self.hook_layer_pre(clip, inp)
self.hook_layer(clip, outp)
class Script(scripts.Script):
def __init__(self):
super().__init__()
self.last_hooker: SDHook|None = None
def title(self):
return NAME
def show(self, is_img2img):
return scripts.AlwaysVisible
def ui(self, is_img2img):
with gr.Accordion(NAME, open=False):
with gr.Row():
enabled = gr.Checkbox(label='Enabled', value=False)
with_hr = gr.Checkbox(label='Also enable on Hires fix', value=False, visible=False)
crop_left = gr.Slider(minimum=-512, maximum=512, step=1, value=0, label='Crop Left')
crop_top = gr.Slider(minimum=-512, maximum=512, step=1, value=0, label='Crop Top')
original_width = gr.Slider(minimum=-1, maximum=4096, step=1, value=-1, label='Original Width (-1 is original size)')
original_height = gr.Slider(minimum=-1, maximum=4096, step=1, value=-1, label='Original Height (-1 is original size)')
target_width = gr.Slider(minimum=-1, maximum=4096, step=1, value=-1, label='Target Width (-1 is original size)')
target_height = gr.Slider(minimum=-1, maximum=4096, step=1, value=-1, label='Target Height (-1 is original size)')
aesthetic_score = gr.Slider(minimum=0.0, maximum=10.0, step=0.05, value=6.0, label="Aesthetic Score (0..10)")
negative_aesthetic_score = gr.Slider(minimum=0.0, maximum=10.0, step=0.05, value=2.5, label="Negative Aesthetic Score (0..10)")
extra_prompt = gr.Textbox(lines=3, label='Extra prompt (set empty to be disabled)')
extra_negative_prompt = gr.Textbox(lines=3, label='Extra negative prompt (set empty to be disabled)')
token_index = gr.Slider(minimum=-77, maximum=76, step=1, value=-1, label='Token index in the prompt for the vector (-1 is first EOT)')
negative_token_index = gr.Slider(minimum=-77, maximum=76, step=1, value=-1, label='Token index in the negative prompt for the vector (-1 is first EOT)')
eot_multiplier = gr.Slider(minimum=-4.0, maximum=8.0, step=0.05, value=1.0, label='Token multiplier')
negative_eot_multiplier = gr.Slider(minimum=-4.0, maximum=8.0, step=0.05, value=1.0, label='Negative token multiplier')
return [
enabled,
crop_left,
crop_top,
original_width,
original_height,
target_width,
target_height,
aesthetic_score,
negative_aesthetic_score,
extra_prompt,
extra_negative_prompt,
token_index,
negative_token_index,
eot_multiplier,
negative_eot_multiplier,
with_hr,
]
def process(
self,
p: StableDiffusionProcessing,
enabled: bool,
crop_left: float,
crop_top: float,
original_width: float,
original_height: float,
target_width: float,
target_height: float,
aesthetic_score: float,
negative_aesthetic_score: float,
extra_prompt: str,
extra_negative_prompt: str,
token_index: float,
negative_token_index: float,
eot_multiplier: float,
negative_eot_multiplier: float,
with_hr: bool,
):
if self.last_hooker is not None:
self.last_hooker.__exit__(None, None, None)
self.last_hooker = None
if not enabled:
return
if original_width < 0:
original_width = p.width
if original_height < 0:
original_height = p.height
if target_width < 0:
target_width = p.width
if target_height < 0:
target_height = p.height
self.last_hooker = Hook(
enabled=True,
p=p,
crop_left=crop_left,
crop_top=crop_top,
original_width=original_width,
original_height=original_height,
target_width=target_width,
target_height=target_height,
aesthetic_score=aesthetic_score,
negative_aesthetic_score=negative_aesthetic_score,
extra_prompt=extra_prompt,
extra_negative_prompt=extra_negative_prompt,
token_index=token_index,
negative_token_index=negative_token_index,
eot_multiplier=eot_multiplier,
negative_eot_multiplier=negative_eot_multiplier,
with_hr=with_hr,
)
self.last_hooker.setup(p)
self.last_hooker.__enter__()
p.extra_generation_params.update({
f'[{NAME}] Enabled': enabled,
#f'[{NAME}] With HR': with_hr,
f'[{NAME}] Crop Left': crop_left,
f'[{NAME}] Crop Top': crop_top,
f'[{NAME}] Original Width': original_width,
f'[{NAME}] Original Height': original_height,
f'[{NAME}] Target Width': target_width,
f'[{NAME}] Target Height': target_height,
f'[{NAME}] Aesthetic Score': aesthetic_score,
f'[{NAME}] Negative Aesthetic Score': negative_aesthetic_score,
f'[{NAME}] Extra Prompt': extra_prompt.__repr__(),
f'[{NAME}] Extra Negative Prompt': extra_negative_prompt.__repr__(),
f'[{NAME}] Token Index': token_index,
f'[{NAME}] Negative Token Index': negative_token_index,
f'[{NAME}] EOT Multiplier': eot_multiplier,
f'[{NAME}] Negative EOT Multiplier': negative_eot_multiplier,
})
if hasattr(p, 'cached_c'):
p.cached_c = [None, None]
if hasattr(p, 'cached_uc'):
p.cached_uc = [None, None]
init_xyz(Script, NAME)
|