sd-webui-agent-scheduler/scripts/task_helpers.py

442 lines
15 KiB
Python

import os
import io
import zlib
import base64
import inspect
import requests
import numpy as np
from enum import Enum
from PIL import Image, ImageOps, ImageChops, ImageEnhance, ImageFilter
from typing import Optional, List
from pydantic import BaseModel, Field
from modules import sd_samplers, scripts
from modules.generation_parameters_copypaste import create_override_settings_dict
from modules.sd_models import CheckpointInfo, get_closet_checkpoint_match
from modules.txt2img import txt2img
from modules.img2img import img2img
from modules.api.models import (
StableDiffusionTxt2ImgProcessingAPI,
StableDiffusionImg2ImgProcessingAPI,
)
from scripts.helpers import log
img2img_image_args_by_mode: dict[int, list[list[str]]] = {
0: [["init_img"]],
1: [["sketch"]],
2: [["init_img_with_mask", "image"], ["init_img_with_mask", "mask"]],
3: [["inpaint_color_sketch"], ["inpaint_color_sketch_orig"]],
4: [["init_img_inpaint"], ["init_mask_inpaint"]],
}
class ControlNetImage(BaseModel):
image: str # base64 or url
mask: Optional[str] = None # base64 or url
class ControlNetUnit(BaseModel):
enabled: Optional[bool] = True
module: Optional[str] = "none"
model: Optional[str] = None
image: ControlNetImage = None
weight: Optional[float] = 1.0
resize_mode: Optional[str] = None
low_vram: Optional[bool] = False
processor_res: Optional[int] = 512
threshold_a: Optional[float] = 64
threshold_b: Optional[float] = 64
guidance_start: Optional[float] = 0.0
guidance_end: Optional[float] = 1.0
pixel_perfect: Optional[bool] = False
control_mode: Optional[str] = "Balanced"
class BaseApiTaskArgs(BaseModel):
task_id: str = Field(exclude=True)
model_hash: str = Field(exclude=True)
prompt: Optional[str] = ""
styles: Optional[List[str]] = []
negative_prompt: Optional[str] = ""
seed: Optional[int] = -1
subseed: Optional[int] = 1
subseed_strength: Optional[int] = 0
seed_resize_from_h: Optional[int] = -1
seed_resize_from_w: Optional[int] = -1
sampler_name: Optional[str] = "DPM++ 2M Karras"
n_iter: Optional[int] = 1
batch_size: Optional[int] = 1
steps: Optional[int] = 20
cfg_scale: Optional[int] = 7.0
restore_faces: Optional[bool] = False
tiling: Optional[bool] = False
width: Optional[int] = 512
height: Optional[int] = 512
script_name: Optional[str] = None
controlnet_args: Optional[List[ControlNetUnit]] = Field(exclude=True, default=[])
override_settings: Optional[dict] = Field(default={})
class Txt2ImgApiTaskArgs(BaseApiTaskArgs):
enable_hr: Optional[bool] = False
denoising_strength: Optional[int] = 0
hr_scale: Optional[int] = 1
hr_upscaler: Optional[str] = "Latent"
hr_second_pass_steps: Optional[int] = 0
hr_resize_x: Optional[int] = 0
hr_resize_y: Optional[int] = 0
class Img2ImgApiTaskArgs(BaseApiTaskArgs):
init_images: List[str]
mask: Optional[str] = None
resize_mode: Optional[int] = 0
denoising_strength: Optional[int] = 0.75
mask_blur: Optional[int] = 4
inpainting_fill: Optional[int] = 0
inpaint_full_res: Optional[bool] = True
inpaint_full_res_padding: Optional[int] = 0
inpainting_mask_invert: Optional[int] = 0
initial_noise_multiplier: Optional[float] = 0.0
def load_image_from_url(url: str):
try:
response = requests.get(url)
buffer = io.BytesIO(response.content)
return Image.open(buffer)
except Exception as e:
log.error(f"[AgentScheduler] Error downloading image from url: {e}")
return None
def load_image(image: str):
if not isinstance(image, str):
return image
pil_image = None
if os.path.exists(image):
pil_image = Image.open(image)
elif image.startswith(("http://", "https://")):
pil_image = load_image_from_url(image)
return pil_image
def load_image_to_nparray(image: str):
pil_image = load_image(image)
return (
np.array(pil_image).astype("uint8")
if isinstance(pil_image, Image.Image)
else None
)
def encode_pil_to_base64(image: Image.Image):
with io.BytesIO() as output_bytes:
image.save(output_bytes, format="PNG")
bytes_data = output_bytes.getvalue()
return base64.b64encode(bytes_data).decode("utf-8")
def load_image_to_base64(image: str):
pil_image = load_image(image)
if not isinstance(pil_image, Image.Image):
return image
return encode_pil_to_base64(pil_image)
def __serialize_image(image):
if isinstance(image, np.ndarray):
shape = image.shape
data = base64.b64encode(zlib.compress(image.tobytes())).decode()
return {"shape": shape, "data": data, "cls": "ndarray"}
elif isinstance(image, Image.Image):
size = image.size
mode = image.mode
data = base64.b64encode(zlib.compress(image.tobytes())).decode()
return {
"size": size,
"mode": mode,
"data": data,
"cls": "Image",
}
else:
return image
def __deserialize_image(image_str):
if isinstance(image_str, dict) and image_str.get("cls", None):
cls = image_str["cls"]
data = zlib.decompress(base64.b64decode(image_str["data"]))
if cls == "ndarray":
shape = tuple(image_str["shape"])
image = np.frombuffer(data, dtype=np.uint8)
return image.reshape(shape)
else:
size = tuple(image_str["size"])
mode = image_str["mode"]
return Image.frombytes(mode, size, data)
else:
return image_str
def serialize_img2img_image_args(args: dict):
for mode, image_args in img2img_image_args_by_mode.items():
for keys in image_args:
if mode != args["mode"]:
# set None to unused image args to save space
args[keys[0]] = None
elif len(keys) == 1:
image = args.get(keys[0], None)
args[keys[0]] = __serialize_image(image)
else:
value = args.get(keys[0], {})
image = value.get(keys[1], None)
value[keys[1]] = __serialize_image(image)
args[keys[0]] = value
def deserialize_img2img_image_args(args: dict):
for mode, image_args in img2img_image_args_by_mode.items():
if mode != args["mode"]:
continue
for keys in image_args:
if len(keys) == 1:
image = args.get(keys[0], None)
args[keys[0]] = __deserialize_image(image)
else:
value = args.get(keys[0], {})
image = value.get(keys[1], None)
value[keys[1]] = __deserialize_image(image)
args[keys[0]] = value
def serialize_controlnet_args(cnet_unit):
args: dict = cnet_unit.__dict__
args["is_cnet"] = True
for k, v in args.items():
if k == "image" and v is not None:
args[k] = {
"image": __serialize_image(v["image"]),
"mask": __serialize_image(v["mask"])
if v.get("mask", None) is not None
else None,
}
if isinstance(v, Enum):
args[k] = v.value
return args
def deserialize_controlnet_args(args: dict):
# args.pop("is_cnet", None)
for k, v in args.items():
if k == "image" and v is not None:
args[k] = {
"image": __deserialize_image(v["image"]),
"mask": __deserialize_image(v["mask"])
if v.get("mask", None) is not None
else None,
}
return args
def map_ui_task_args_list_to_named_args(
args: list, is_img2img: bool, checkpoint: str = None
):
args_name = []
if is_img2img:
args_name = inspect.getfullargspec(img2img).args
else:
args_name = inspect.getfullargspec(txt2img).args
named_args = dict(zip(args_name, args[0 : len(args_name)]))
script_args = args[len(args_name) :]
if checkpoint is not None:
override_settings_texts = named_args.get("override_settings_texts", [])
override_settings_texts.append("Model hash: " + checkpoint)
named_args["override_settings_texts"] = override_settings_texts
return (
named_args,
script_args,
)
def map_ui_task_args_to_api_task_args(
named_args: dict, script_args: list, is_img2img: bool
):
api_task_args: dict = named_args.copy()
prompt_styles = api_task_args.pop("prompt_styles", [])
api_task_args["styles"] = prompt_styles
sampler_index = api_task_args.pop("sampler_index", 0)
api_task_args["sampler_name"] = sd_samplers.samplers[sampler_index].name
override_settings_texts = api_task_args.pop("override_settings_texts", [])
api_task_args["override_settings"] = create_override_settings_dict(
override_settings_texts
)
if is_img2img:
mode = api_task_args.pop("mode", 0)
for arg_mode, image_args in img2img_image_args_by_mode.items():
if mode != arg_mode:
for keys in image_args:
api_task_args.pop(keys[0], None)
# the logic below is copied from modules/img2img.py
if mode == 0:
image = api_task_args.pop("init_img").convert("RGB")
mask = None
elif mode == 1:
image = api_task_args.pop("sketch").convert("RGB")
mask = None
elif mode == 2:
init_img_with_mask: dict = api_task_args.pop("init_img_with_mask")
image = init_img_with_mask.get("image").convert("RGB")
mask = init_img_with_mask.get("mask")
alpha_mask = (
ImageOps.invert(image.split()[-1])
.convert("L")
.point(lambda x: 255 if x > 0 else 0, mode="1")
)
mask = ImageChops.lighter(alpha_mask, mask.convert("L")).convert("L")
elif mode == 3:
image = api_task_args.pop("inpaint_color_sketch")
orig = api_task_args.pop("inpaint_color_sketch_orig") or image
mask_alpha = api_task_args.pop("mask_alpha", 0)
mask_blur = api_task_args.get("mask_blur", 4)
pred = np.any(np.array(image) != np.array(orig), axis=-1)
mask = Image.fromarray(pred.astype(np.uint8) * 255, "L")
mask = ImageEnhance.Brightness(mask).enhance(1 - mask_alpha / 100)
blur = ImageFilter.GaussianBlur(mask_blur)
image = Image.composite(image.filter(blur), orig, mask.filter(blur))
image = image.convert("RGB")
elif mode == 4:
image = api_task_args.pop("init_img_inpaint")
mask = api_task_args.pop("init_mask_inpaint")
else:
raise Exception(f"Batch mode is not supported yet")
image = ImageOps.exif_transpose(image)
api_task_args["init_images"] = [encode_pil_to_base64(image)]
api_task_args["mask"] = encode_pil_to_base64(mask) if mask is not None else None
selected_scale_tab = api_task_args.pop("selected_scale_tab", 0)
scale_by = api_task_args.pop("scale_by", 1)
if selected_scale_tab == 1:
api_task_args["width"] = int(image.width * scale_by)
api_task_args["height"] = int(image.height * scale_by)
else:
hr_sampler_index = api_task_args.pop("hr_sampler_index", 0)
api_task_args["hr_sampler_name"] = (
sd_samplers.samplers_for_img2img[hr_sampler_index - 1].name
if hr_sampler_index != 0
else None
)
# script
script_runner = scripts.scripts_img2img if is_img2img else scripts.scripts_txt2img
script_id = script_args[0]
if script_id == 0:
api_task_args["script_name"] = None
api_task_args["script_args"] = []
else:
script = script_runner.selectable_scripts[script_id - 1]
api_task_args["script_name"] = script.title.lower()
api_task_args["script_args"] = script_args[script.args_from : script.args_to]
# alwayson scripts
alwayson_scripts = api_task_args.get("alwayson_scripts", None)
if not alwayson_scripts or not isinstance(alwayson_scripts, dict):
alwayson_scripts = {}
api_task_args["alwayson_scripts"] = alwayson_scripts
for script in script_runner.alwayson_scripts:
alwayson_script_args = script_args[script.args_from : script.args_to]
if script.title.lower() == "controlnet":
for i, cnet_args in enumerate(alwayson_script_args):
alwayson_script_args[i] = serialize_controlnet_args(cnet_args)
alwayson_scripts[script.title.lower()] = {"args": alwayson_script_args}
return api_task_args
def serialize_api_task_args(params: dict, is_img2img: bool):
# pop out custom params
model_hash = params.pop("model_hash", None)
controlnet_args = params.pop("controlnet_args", None)
args = (
StableDiffusionImg2ImgProcessingAPI(**params)
if is_img2img
else StableDiffusionTxt2ImgProcessingAPI(**params)
)
if args.override_settings is None:
args.override_settings = {}
if model_hash is None:
model_hash = args.override_settings.get("sd_model_checkpoint", None)
if model_hash is None:
log.error("[AgentScheduler] API task must supply model hash")
return
checkpoint: CheckpointInfo = get_closet_checkpoint_match(model_hash)
if not checkpoint:
log.warn(f"[AgentScheduler] No checkpoint found for model hash {model_hash}")
return
args.override_settings["sd_model_checkpoint"] = checkpoint.title
# load images from url or file if needed
if is_img2img:
init_images = args.init_images
for i, image in enumerate(init_images):
init_images[i] = load_image_to_base64(image)
args.mask = load_image_to_base64(args.mask)
# handle custom controlnet args
if controlnet_args is not None:
if args.alwayson_scripts is None:
args.alwayson_scripts = {}
controlnets = []
for cnet in controlnet_args:
enabled = cnet.get("enabled", True)
cnet_image = cnet.get("image", None)
if not enabled:
continue
if not isinstance(cnet_image, dict):
log.error(f"[AgentScheduler] Controlnet image is required")
continue
image = cnet_image.get("image", None)
mask = cnet_image.get("mask", None)
if image is None:
log.error(f"[AgentScheduler] Controlnet image is required")
continue
# load controlnet images from url or file if needed
cnet_image["image"] = load_image_to_base64(image)
cnet_image["mask"] = load_image_to_base64(mask)
controlnets.append(cnet)
if len(controlnets) > 0:
args.alwayson_scripts["controlnet"] = {"args": controlnets}
return args.dict()