Files
THtianhao-ComfyUI-FaceChain/facechain/style_loader_node.py
T
2024-01-05 18:49:40 +08:00

189 lines
7.1 KiB
Python

import os
import json
import shutil
import comfy.utils
import comfy
import comfy.sd
import modelscope
# from facechain.constants import neg_prompt as neg, pos_prompt_with_cloth, pos_prompt_with_style, \
# pose_models, pose_examples, base_models, tts_speakers_map
#from facechain.utils import snapshot_download, check_ffmpeg, set_spawn_method, project_dir, join_worker_data_dir
from modelscope import snapshot_download
import folder_paths
import sys
my_dir = os.path.dirname(os.path.abspath(__file__))
custom_nodes_dir = os.path.abspath(os.path.join(my_dir, '../../ComfyUI-FaceChain'))
#comfy_dir = os.path.abspath(os.path.join(my_dir, '..', '..'))
comfy_dir = os.path.abspath(os.path.join(my_dir, '../..'))
# Append comfy_dir to sys.path & import files
sys.path.append(comfy_dir)
#all_data = []
# 名字到数据的映射
name_map = {}
style_list = []
base_models = [
{'name': 'leosamsMoonfilm_filmGrain20',
'model_id': 'ly261666/cv_portrait_model',
'revision': 'v2.0',
'sub_path': "film/film"},
{'name': 'MajicmixRealistic_v6',
'model_id': 'YorickHe/majicmixRealistic_v6',
'revision': 'v1.0.0',
'sub_path': "realistic"},
]
class FCStyleLoraLoad:
def __init__(self):
self.loaded_lora = None
print("FC StyleLoraLoad initing")
# 存储所有数据
load_style_files()
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"style_name": (load_style_files(),),
}
}
RETURN_TYPES = ("MODEL", "CLIP", "VAE", "STRING")
FUNCTION = "style_lora_load"
CATEGORY = "facechain/lora"
def style_lora_load(self, model=None, clip=None, style_name=None):
style_data = get_data_by_name(style_name)
base_model_name = style_data["base_model_name"]
matching_model = next((model for model in base_models if model['name'] == base_model_name), None)
base_model = matching_model['model_id']
base_model_revision = matching_model['revision']
base_model_sub_path = matching_model['sub_path']
style_model_id = style_data["model_id"]
style_revision = style_data["revision"]
add_prompt_style = style_data["add_prompt_style"]
style_bin_file = style_data["bin_file"]
style_multiplier_style = style_data["multiplier_style"]
style_multiplier_human = style_data["multiplier_human"]
# matched = list(filter(lambda item: style_name == item['name'], styles))
# if len(matched) == 0:
# raise ValueError(f'styles not found: {style_name}')
# matched = matched[0]
# style_model = matched['name']
if style_model_id is None:
style_model_path = None
else:
model_dir = snapshot_download(style_model_id, style_revision)
style_model_path = os.path.join(model_dir, style_bin_file)
#if using modelscope, use following code download model
if matching_model['name'] == 'leosamsMoonfilm_filmGrain20':
base_model_name = 'leosamsMoonfilm_filmGrain20.safetensors'
if matching_model['name'] == 'majicmixRealistic_v6':
base_model_name = 'majicmixRealistic_v6.safetensors'
ckpt_path = folder_paths.get_full_path("checkpoints", base_model_name)
#check if modelfile not exist, download it from modelscope
if ckpt_path == None:
model_cache_dir = snapshot_download(f'ultimatech/{matching_model["name"]}')
# shutil.move(os.path.join(model_cache_dir,base_model_name),folder_paths.folder_names_and_paths["checkpoints"][0][0])
# shutil.rmtree(model_cache_dir)
ckpt_path = os.path.join(model_cache_dir,base_model_name)
out = comfy.sd.load_checkpoint_guess_config(ckpt_path, output_vae=True, output_clip=True,
embedding_directory=folder_paths.get_folder_paths("embeddings"))
model_patcher, clip, vae, clipvision = out
if style_model_id is None:
model_dir = snapshot_download('Cherrytest/zjz_mj_jiyi_small_addtxt_fromleo', revision='v1.0.0')
style_model_path = os.path.join(model_dir, 'zjz_mj_jiyi_small_addtxt_fromleo.safetensors')
lora = comfy.utils.load_torch_file(style_model_path, safe_load=True)
model_lora, clip_lora = comfy.sd.load_lora_for_models(model_patcher, clip, lora, style_multiplier_style, 1)
return (model_lora, clip_lora, vae, add_prompt_style)
else:
lora = comfy.utils.load_torch_file(style_model_path, safe_load=True)
model_lora, clip_lora = comfy.sd.load_lora_for_models(model_patcher, clip, lora, style_multiplier_style, 1)
return(model_lora, clip_lora, vae, add_prompt_style)
def load_style_files(self):
folder = "styles"
for subdir in os.listdir(folder):
subdir_path = os.path.join(folder, subdir)
for json_file in os.listdir(subdir_path):
json_path = os.path.join(subdir_path, json_file)
with open(json_path, encoding='utf-8') as f:
json_data = json.load(f)
#all_data.append(json_data)
name = json_data['name']
name_map[name] = json_data
# style_list = json.dumps(dict(name_map.keys()))
# print(style_list)
return list(name_map.keys())
# 根据名称查询
# 读取文件和解析
def load_style_files():
folder = f"{custom_nodes_dir}/styles"
for subdir in os.listdir(folder):
subdir_path = os.path.join(folder, subdir)
for json_file in os.listdir(subdir_path):
json_path = os.path.join(subdir_path, json_file)
with open(json_path, encoding='utf-8') as f:
json_data = json.load(f)
json_data["base_model_name"] = subdir
#all_data.append(json_data)
name = json_data['name']
name_map[name] = json_data
style_list = list(name_map.keys())
#print(style_list)
return style_list
NODE_CLASS_MAPPINGS = {
"FCStyleLoraLoad": FCStyleLoraLoad,
}
def get_data_by_name(name):
return name_map[name]
styles = []
# for base_model in base_models:
# style_in_base = []
# folder_path = f"{os.path.dirname(os.path.abspath(__file__))}/styles/{base_model['name']}"
# files = os.listdir(folder_path)
# files.sort()
# for file in files:
# file_path = os.path.join(folder_path, file)
# with open(file_path, "r", encoding='utf-8') as f:
# data = json.load(f)
# # if data['img'][:2] == './':
# # data['img'] = f"{project_dir}/{data['img'][2:]}"
# style_in_base.append(data['name'])
# styles.append(data)
# base_model['style_list'] = style_in_base
def main():
folder = 'styles'
# 加载文件
fc = FCStyleLoraLoad();
#fc.load_style_files()
fc.style_lora_load("","",'盔甲风(Armor)')
fc.style_lora_load("", "", "秋日胡杨风(Autumn populus euphratica style)")
# 测试根据名称查询
name = "秋日胡杨风(Autumn populus euphratica style)"
print(fc.name_map[name])
if __name__ == '__main__':
main()