hugging-apps/echo-memory

1from typing_extensions import Literal, TypeAlias2 3from ..models.sd_text_encoder import SDTextEncoder4from ..models.sd_unet import SDUNet5from ..models.sd_vae_encoder import SDVAEEncoder6from ..models.sd_vae_decoder import SDVAEDecoder7 8from ..models.sdxl_text_encoder import SDXLTextEncoder, SDXLTextEncoder29from ..models.sdxl_unet import SDXLUNet10from ..models.sdxl_vae_decoder import SDXLVAEDecoder11from ..models.sdxl_vae_encoder import SDXLVAEEncoder12 13from ..models.sd3_text_encoder import SD3TextEncoder1, SD3TextEncoder2, SD3TextEncoder314from ..models.sd3_dit import SD3DiT15from ..models.sd3_vae_decoder import SD3VAEDecoder16from ..models.sd3_vae_encoder import SD3VAEEncoder17 18from ..models.sd_controlnet import SDControlNet19from ..models.sdxl_controlnet import SDXLControlNetUnion20 21from ..models.sd_motion import SDMotionModel22from ..models.sdxl_motion import SDXLMotionModel23 24from ..models.svd_image_encoder import SVDImageEncoder25from ..models.svd_unet import SVDUNet26from ..models.svd_vae_decoder import SVDVAEDecoder27from ..models.svd_vae_encoder import SVDVAEEncoder28 29from ..models.sd_ipadapter import SDIpAdapter, IpAdapterCLIPImageEmbedder30from ..models.sdxl_ipadapter import SDXLIpAdapter, IpAdapterXLCLIPImageEmbedder31 32from ..models.hunyuan_dit_text_encoder import HunyuanDiTCLIPTextEncoder, HunyuanDiTT5TextEncoder33from ..models.hunyuan_dit import HunyuanDiT34 35from ..models.flux_dit import FluxDiT36from ..models.flux_text_encoder import FluxTextEncoder237from ..models.flux_vae import FluxVAEEncoder, FluxVAEDecoder38from ..models.flux_controlnet import FluxControlNet39from ..models.flux_ipadapter import FluxIpAdapter40from ..models.flux_infiniteyou import InfiniteYouImageProjector41 42from ..models.cog_vae import CogVAEEncoder, CogVAEDecoder43from ..models.cog_dit import CogDiT44 45from ..models.omnigen import OmniGenTransformer46 47from ..models.hunyuan_video_vae_decoder import HunyuanVideoVAEDecoder48from ..models.hunyuan_video_vae_encoder import HunyuanVideoVAEEncoder49 50from ..extensions.RIFE import IFNet51from ..extensions.ESRGAN import RRDBNet52 53from ..models.hunyuan_video_dit import HunyuanVideoDiT54 55from ..models.stepvideo_vae import StepVideoVAE56from ..models.stepvideo_dit import StepVideoModel57 58from ..models.wan_video_dit import WanModel59from ..models.wan_video_text_encoder import WanTextEncoder60from ..models.wan_video_image_encoder import WanImageEncoder61from ..models.wan_video_vae import WanVideoVAE62from ..models.wan_video_motion_controller import WanMotionControllerModel63from ..models.wan_video_vace import VaceWanModel64 65from ..models.step1x_connector import Qwen2Connector66 67from ..models.flux_value_control import SingleValueEncoder68 69from ..lora.flux_lora import FluxLoraPatcher70from ..models.flux_lora_encoder import FluxLoRAEncoder71 72 73model_loader_configs = [74    # These configs are provided for detecting model type automatically.75    # The format is (state_dict_keys_hash, state_dict_keys_hash_with_shape, model_names, model_classes, model_resource)76    (None, "091b0e30e77c76626b3ba62acdf95343", ["sd_controlnet"], [SDControlNet], "civitai"),77    (None, "4a6c8306a27d916dea81263c8c88f450", ["hunyuan_dit_clip_text_encoder"], [HunyuanDiTCLIPTextEncoder], "civitai"),78    (None, "f4aec400fe394297961218c768004521", ["hunyuan_dit"], [HunyuanDiT], "civitai"),79    (None, "9e6e58043a5a2e332803ed42f6ee7181", ["hunyuan_dit_t5_text_encoder"], [HunyuanDiTT5TextEncoder], "civitai"),80    (None, "13115dd45a6e1c39860f91ab073b8a78", ["sdxl_vae_encoder", "sdxl_vae_decoder"], [SDXLVAEEncoder, SDXLVAEDecoder], "diffusers"),81    (None, "d78aa6797382a6d455362358a3295ea9", ["sd_ipadapter_clip_image_encoder"], [IpAdapterCLIPImageEmbedder], "diffusers"),82    (None, "e291636cc15e803186b47404262ef812", ["sd_ipadapter"], [SDIpAdapter], "civitai"),83    (None, "399c81f2f8de8d1843d0127a00f3c224", ["sdxl_ipadapter_clip_image_encoder"], [IpAdapterXLCLIPImageEmbedder], "diffusers"),84    (None, "a64eac9aa0db4b9602213bc0131281c7", ["sdxl_ipadapter"], [SDXLIpAdapter], "civitai"),85    (None, "52817e4fdd89df154f02749ca6f692ac", ["sdxl_unet"], [SDXLUNet], "diffusers"),86    (None, "03343c606f16d834d6411d0902b53636", ["sd_text_encoder", "sd_unet", "sd_vae_decoder", "sd_vae_encoder"], [SDTextEncoder, SDUNet, SDVAEDecoder, SDVAEEncoder], "civitai"),87    (None, "d4ba77a7ece070679b4a987f58f201e9", ["sd_text_encoder"], [SDTextEncoder], "civitai"),88    (None, "d0c89e55c5a57cf3981def0cb1c9e65a", ["sd_vae_decoder", "sd_vae_encoder"], [SDVAEDecoder, SDVAEEncoder], "civitai"),89    (None, "3926bf373b39a67eeafd7901478a47a7", ["sd_unet"], [SDUNet], "civitai"),90    (None, "1e0c39ec176b9007c05f76d52b554a4d", ["sd3_text_encoder_1", "sd3_text_encoder_2", "sd3_dit", "sd3_vae_encoder", "sd3_vae_decoder"], [SD3TextEncoder1, SD3TextEncoder2, SD3DiT, SD3VAEEncoder, SD3VAEDecoder], "civitai"),91    (None, "d9e0290829ba8d98e28e1a2b1407db4a", ["sd3_text_encoder_1", "sd3_text_encoder_2", "sd3_text_encoder_3", "sd3_dit", "sd3_vae_encoder", "sd3_vae_decoder"], [SD3TextEncoder1, SD3TextEncoder2, SD3TextEncoder3, SD3DiT, SD3VAEEncoder, SD3VAEDecoder], "civitai"),92    (None, "5072d0b24e406b49507abe861cf97691", ["sd3_text_encoder_3"], [SD3TextEncoder3], "civitai"),93    (None, "4cf64a799d04260df438c6f33c9a047e", ["sdxl_text_encoder", "sdxl_text_encoder_2", "sdxl_unet", "sdxl_vae_decoder", "sdxl_vae_encoder"], [SDXLTextEncoder, SDXLTextEncoder2, SDXLUNet, SDXLVAEDecoder, SDXLVAEEncoder], "civitai"),94    (None, "d9b008a867c498ab12ad24042eff8e3f", ["sdxl_text_encoder", "sdxl_text_encoder_2", "sdxl_unet", "sdxl_vae_decoder", "sdxl_vae_encoder"], [SDXLTextEncoder, SDXLTextEncoder2, SDXLUNet, SDXLVAEDecoder, SDXLVAEEncoder], "civitai"), # SDXL-Turbo95    (None, "025bb7452e531a3853d951d77c63f032", ["sdxl_text_encoder", "sdxl_text_encoder_2"], [SDXLTextEncoder, SDXLTextEncoder2], "civitai"),96    (None, "298997b403a4245c04102c9f36aac348", ["sdxl_unet"], [SDXLUNet], "civitai"),97    (None, "2a07abce74b4bdc696b76254ab474da6", ["svd_image_encoder", "svd_unet", "svd_vae_decoder", "svd_vae_encoder"], [SVDImageEncoder, SVDUNet, SVDVAEDecoder, SVDVAEEncoder], "civitai"),98    (None, "c96a285a6888465f87de22a984d049fb", ["sd_motion_modules"], [SDMotionModel], "civitai"),99    (None, "72907b92caed19bdb2adb89aa4063fe2", ["sdxl_motion_modules"], [SDXLMotionModel], "civitai"),100    (None, "31d2d9614fba60511fc9bf2604aa01f7", ["sdxl_controlnet"], [SDXLControlNetUnion], "diffusers"),101    (None, "94eefa3dac9cec93cb1ebaf1747d7b78", ["sd3_text_encoder_1"], [SD3TextEncoder1], "diffusers"),102    (None, "1aafa3cc91716fb6b300cc1cd51b85a3", ["flux_vae_encoder", "flux_vae_decoder"], [FluxVAEEncoder, FluxVAEDecoder], "diffusers"),103    (None, "21ea55f476dfc4fd135587abb59dfe5d", ["flux_vae_encoder", "flux_vae_decoder"], [FluxVAEEncoder, FluxVAEDecoder], "civitai"),104    (None, "a29710fea6dddb0314663ee823598e50", ["flux_dit"], [FluxDiT], "civitai"),105    (None, "57b02550baab820169365b3ee3afa2c9", ["flux_dit"], [FluxDiT], "civitai"),106    (None, "3394f306c4cbf04334b712bf5aaed95f", ["flux_dit"], [FluxDiT], "civitai"),107    (None, "023f054d918a84ccf503481fd1e3379e", ["flux_dit"], [FluxDiT], "civitai"),108    (None, "d02f41c13549fa5093d3521f62a5570a", ["flux_dit"], [FluxDiT], "civitai"),109    (None, "605c56eab23e9e2af863ad8f0813a25d", ["flux_dit"], [FluxDiT], "diffusers"),110    (None, "0629116fce1472503a66992f96f3eb1a", ["flux_value_controller"], [SingleValueEncoder], "civitai"),111    (None, "280189ee084bca10f70907bf6ce1649d", ["cog_vae_encoder", "cog_vae_decoder"], [CogVAEEncoder, CogVAEDecoder], "diffusers"),112    (None, "9b9313d104ac4df27991352fec013fd4", ["rife"], [IFNet], "civitai"),113    (None, "6b7116078c4170bfbeaedc8fe71f6649", ["esrgan"], [RRDBNet], "civitai"),114    (None, "61cbcbc7ac11f169c5949223efa960d1", ["omnigen_transformer"], [OmniGenTransformer], "diffusers"),115    (None, "78d18b9101345ff695f312e7e62538c0", ["flux_controlnet"], [FluxControlNet], "diffusers"),116    (None, "b001c89139b5f053c715fe772362dd2a", ["flux_controlnet"], [FluxControlNet], "diffusers"),117    (None, "52357cb26250681367488a8954c271e8", ["flux_controlnet"], [FluxControlNet], "diffusers"),118    (None, "0cfd1740758423a2a854d67c136d1e8c", ["flux_controlnet"], [FluxControlNet], "diffusers"),119    (None, "7f9583eb8ba86642abb9a21a4b2c9e16", ["flux_controlnet"], [FluxControlNet], "diffusers"),120    (None, "43ad5aaa27dd4ee01b832ed16773fa52", ["flux_controlnet"], [FluxControlNet], "diffusers"),121    (None, "c07c0f04f5ff55e86b4e937c7a40d481", ["infiniteyou_image_projector"], [InfiniteYouImageProjector], "diffusers"),122    (None, "4daaa66cc656a8fe369908693dad0a35", ["flux_ipadapter"], [FluxIpAdapter], "diffusers"),123    (None, "51aed3d27d482fceb5e0739b03060e8f", ["sd3_dit", "sd3_vae_encoder", "sd3_vae_decoder"], [SD3DiT, SD3VAEEncoder, SD3VAEDecoder], "civitai"),124    (None, "98cc34ccc5b54ae0e56bdea8688dcd5a", ["sd3_text_encoder_2"], [SD3TextEncoder2], "civitai"),125    (None, "77ff18050dbc23f50382e45d51a779fe", ["sd3_dit", "sd3_vae_encoder", "sd3_vae_decoder"], [SD3DiT, SD3VAEEncoder, SD3VAEDecoder], "civitai"),126    (None, "5da81baee73198a7c19e6d2fe8b5148e", ["sd3_text_encoder_1"], [SD3TextEncoder1], "diffusers"),127    (None, "aeb82dce778a03dcb4d726cb03f3c43f", ["hunyuan_video_vae_decoder", "hunyuan_video_vae_encoder"], [HunyuanVideoVAEDecoder, HunyuanVideoVAEEncoder], "diffusers"),128    (None, "b9588f02e78f5ccafc9d7c0294e46308", ["hunyuan_video_dit"], [HunyuanVideoDiT], "civitai"),129    (None, "84ef4bd4757f60e906b54aa6a7815dc6", ["hunyuan_video_dit"], [HunyuanVideoDiT], "civitai"),130    (None, "68beaf8429b7c11aa8ca05b1bd0058bd", ["stepvideo_vae"], [StepVideoVAE], "civitai"),131    (None, "5c0216a2132b082c10cb7a0e0377e681", ["stepvideo_dit"], [StepVideoModel], "civitai"),132    (None, "9269f8db9040a9d860eaca435be61814", ["wan_video_dit"], [WanModel], "civitai"),133    (None, "aafcfd9672c3a2456dc46e1cb6e52c70", ["wan_video_dit"], [WanModel], "civitai"),134    (None, "6bfcfb3b342cb286ce886889d519a77e", ["wan_video_dit"], [WanModel], "civitai"),135    (None, "6d6ccde6845b95ad9114ab993d917893", ["wan_video_dit"], [WanModel], "civitai"),136    (None, "6bfcfb3b342cb286ce886889d519a77e", ["wan_video_dit"], [WanModel], "civitai"),137    (None, "349723183fc063b2bfc10bb2835cf677", ["wan_video_dit"], [WanModel], "civitai"),138    (None, "efa44cddf936c70abd0ea28b6cbe946c", ["wan_video_dit"], [WanModel], "civitai"),139    (None, "3ef3b1f8e1dab83d5b71fd7b617f859f", ["wan_video_dit"], [WanModel], "civitai"),140    (None, "70ddad9d3a133785da5ea371aae09504", ["wan_video_dit"], [WanModel], "civitai"),141    (None, "26bde73488a92e64cc20b0a7485b9e5b", ["wan_video_dit"], [WanModel], "civitai"),142    (None, "ac6a5aa74f4a0aab6f64eb9a72f19901", ["wan_video_dit"], [WanModel], "civitai"), 143    (None, "b61c605c2adbd23124d152ed28e049ae", ["wan_video_dit"], [WanModel], "civitai"), 144    (None, "a61453409b67cd3246cf0c3bebad47ba", ["wan_video_dit", "wan_video_vace"], [WanModel, VaceWanModel], "civitai"),145    (None, "7a513e1f257a861512b1afd387a8ecd9", ["wan_video_dit", "wan_video_vace"], [WanModel, VaceWanModel], "civitai"),146    (None, "cb104773c6c2cb6df4f9529ad5c60d0b", ["wan_video_dit"], [WanModel], "diffusers"),147    # 添加合并后带 action 模块的模型 hash (stage-2-key-14B-merged)148    (None, "3296967e8209bf52ffdb9b8414dab595", ["wan_video_dit"], [WanModel], "civitai"),  # 合并后带 action 模块的模型（正确的hash）149    (None, "9c8818c2cbea55eca56c7b447df170da", ["wan_video_text_encoder"], [WanTextEncoder], "civitai"),150    (None, "5941c53e207d62f20f9025686193c40b", ["wan_video_image_encoder"], [WanImageEncoder], "civitai"),151    (None, "1378ea763357eea97acdef78e65d6d96", ["wan_video_vae"], [WanVideoVAE], "civitai"),152    (None, "ccc42284ea13e1ad04693284c7a09be6", ["wan_video_vae"], [WanVideoVAE], "civitai"),153    (None, "dbd5ec76bbf977983f972c151d545389", ["wan_video_motion_controller"], [WanMotionControllerModel], "civitai"),154    (None, "d30fb9e02b1dbf4e509142f05cf7dd50", ["flux_dit", "step1x_connector"], [FluxDiT, Qwen2Connector], "civitai"),155    (None, "30143afb2dea73d1ac580e0787628f8c", ["flux_lora_patcher"], [FluxLoraPatcher], "civitai"),156    (None, "77c2e4dd2440269eb33bfaa0d004f6ab", ["flux_lora_encoder"], [FluxLoRAEncoder], "civitai"),157]158huggingface_model_loader_configs = [159    # These configs are provided for detecting model type automatically.160    # The format is (architecture_in_huggingface_config, huggingface_lib, model_name, redirected_architecture)161    ("ChatGLMModel", "diffsynth.models.kolors_text_encoder", "kolors_text_encoder", None),162    ("MarianMTModel", "transformers.models.marian.modeling_marian", "translator", None),163    ("BloomForCausalLM", "transformers.models.bloom.modeling_bloom", "beautiful_prompt", None),164    ("Qwen2ForCausalLM", "transformers.models.qwen2.modeling_qwen2", "qwen_prompt", None),165    # ("LlamaForCausalLM", "transformers.models.llama.modeling_llama", "omost_prompt", None),166    ("T5EncoderModel", "diffsynth.models.flux_text_encoder", "flux_text_encoder_2", "FluxTextEncoder2"),167    ("CogVideoXTransformer3DModel", "diffsynth.models.cog_dit", "cog_dit", "CogDiT"),168    ("SiglipModel", "transformers.models.siglip.modeling_siglip", "siglip_vision_model", "SiglipVisionModel"),169    ("LlamaForCausalLM", "diffsynth.models.hunyuan_video_text_encoder", "hunyuan_video_text_encoder_2", "HunyuanVideoLLMEncoder"),170    ("LlavaForConditionalGeneration", "diffsynth.models.hunyuan_video_text_encoder", "hunyuan_video_text_encoder_2", "HunyuanVideoMLLMEncoder"),171    ("Step1Model", "diffsynth.models.stepvideo_text_encoder", "stepvideo_text_encoder_2", "STEP1TextEncoder"),172    ("Qwen2_5_VLForConditionalGeneration", "diffsynth.models.qwenvl", "qwenvl", "Qwen25VL_7b_Embedder"),173]174patch_model_loader_configs = [175    # These configs are provided for detecting model type automatically.176    # The format is (state_dict_keys_hash_with_shape, model_name, model_class, extra_kwargs)177    ("9a4ab6869ac9b7d6e31f9854e397c867", ["svd_unet"], [SVDUNet], {"add_positional_conv": 128}),178]179 180preset_models_on_huggingface = {181    "HunyuanDiT": [182        ("Tencent-Hunyuan/HunyuanDiT", "t2i/clip_text_encoder/pytorch_model.bin", "models/HunyuanDiT/t2i/clip_text_encoder"),183        ("Tencent-Hunyuan/HunyuanDiT", "t2i/mt5/pytorch_model.bin", "models/HunyuanDiT/t2i/mt5"),184        ("Tencent-Hunyuan/HunyuanDiT", "t2i/model/pytorch_model_ema.pt", "models/HunyuanDiT/t2i/model"),185        ("Tencent-Hunyuan/HunyuanDiT", "t2i/sdxl-vae-fp16-fix/diffusion_pytorch_model.bin", "models/HunyuanDiT/t2i/sdxl-vae-fp16-fix"),186    ],187    "stable-video-diffusion-img2vid-xt": [188        ("stabilityai/stable-video-diffusion-img2vid-xt", "svd_xt.safetensors", "models/stable_video_diffusion"),189    ],190    "ExVideo-SVD-128f-v1": [191        ("ECNU-CILab/ExVideo-SVD-128f-v1", "model.fp16.safetensors", "models/stable_video_diffusion"),192    ],193    # Stable Diffusion194    "StableDiffusion_v15": [195        ("benjamin-paine/stable-diffusion-v1-5", "v1-5-pruned-emaonly.safetensors", "models/stable_diffusion"),196    ],197    "DreamShaper_8": [198        ("Yntec/Dreamshaper8", "dreamshaper_8.safetensors", "models/stable_diffusion"),199    ],200    # Textual Inversion201    "TextualInversion_VeryBadImageNegative_v1.3": [202        ("gemasai/verybadimagenegative_v1.3", "verybadimagenegative_v1.3.pt", "models/textual_inversion"),203    ],204    # Stable Diffusion XL205    "StableDiffusionXL_v1": [206        ("stabilityai/stable-diffusion-xl-base-1.0", "sd_xl_base_1.0.safetensors", "models/stable_diffusion_xl"),207    ],208    "BluePencilXL_v200": [209        ("frankjoshua/bluePencilXL_v200", "bluePencilXL_v200.safetensors", "models/stable_diffusion_xl"),210    ],211    "StableDiffusionXL_Turbo": [212        ("stabilityai/sdxl-turbo", "sd_xl_turbo_1.0_fp16.safetensors", "models/stable_diffusion_xl_turbo"),213    ],214    # Stable Diffusion 3215    "StableDiffusion3": [216        ("stabilityai/stable-diffusion-3-medium", "sd3_medium_incl_clips_t5xxlfp16.safetensors", "models/stable_diffusion_3"),217    ],218    "StableDiffusion3_without_T5": [219        ("stabilityai/stable-diffusion-3-medium", "sd3_medium_incl_clips.safetensors", "models/stable_diffusion_3"),220    ],221    # ControlNet222    "ControlNet_v11f1p_sd15_depth": [223        ("lllyasviel/ControlNet-v1-1", "control_v11f1p_sd15_depth.pth", "models/ControlNet"),224        ("lllyasviel/Annotators", "dpt_hybrid-midas-501f0c75.pt", "models/Annotators")225    ],226    "ControlNet_v11p_sd15_softedge": [227        ("lllyasviel/ControlNet-v1-1", "control_v11p_sd15_softedge.pth", "models/ControlNet"),228        ("lllyasviel/Annotators", "ControlNetHED.pth", "models/Annotators")229    ],230    "ControlNet_v11f1e_sd15_tile": [231        ("lllyasviel/ControlNet-v1-1", "control_v11f1e_sd15_tile.pth", "models/ControlNet")232    ],233    "ControlNet_v11p_sd15_lineart": [234        ("lllyasviel/ControlNet-v1-1", "control_v11p_sd15_lineart.pth", "models/ControlNet"),235        ("lllyasviel/Annotators", "sk_model.pth", "models/Annotators"),236        ("lllyasviel/Annotators", "sk_model2.pth", "models/Annotators")237    ],238    "ControlNet_union_sdxl_promax": [239        ("xinsir/controlnet-union-sdxl-1.0", "diffusion_pytorch_model_promax.safetensors", "models/ControlNet/controlnet_union"),240        ("lllyasviel/Annotators", "dpt_hybrid-midas-501f0c75.pt", "models/Annotators")241    ],242    # AnimateDiff243    "AnimateDiff_v2": [244        ("guoyww/animatediff", "mm_sd_v15_v2.ckpt", "models/AnimateDiff"),245    ],246    "AnimateDiff_xl_beta": [247        ("guoyww/animatediff", "mm_sdxl_v10_beta.ckpt", "models/AnimateDiff"),248    ],249 250    # Qwen Prompt251    "QwenPrompt": [252        ("Qwen/Qwen2-1.5B-Instruct", "config.json", "models/QwenPrompt/qwen2-1.5b-instruct"),253        ("Qwen/Qwen2-1.5B-Instruct", "generation_config.json", "models/QwenPrompt/qwen2-1.5b-instruct"),254        ("Qwen/Qwen2-1.5B-Instruct", "model.safetensors", "models/QwenPrompt/qwen2-1.5b-instruct"),255        ("Qwen/Qwen2-1.5B-Instruct", "special_tokens_map.json", "models/QwenPrompt/qwen2-1.5b-instruct"),256        ("Qwen/Qwen2-1.5B-Instruct", "tokenizer.json", "models/QwenPrompt/qwen2-1.5b-instruct"),257        ("Qwen/Qwen2-1.5B-Instruct", "tokenizer_config.json", "models/QwenPrompt/qwen2-1.5b-instruct"),258        ("Qwen/Qwen2-1.5B-Instruct", "merges.txt", "models/QwenPrompt/qwen2-1.5b-instruct"),259        ("Qwen/Qwen2-1.5B-Instruct", "vocab.json", "models/QwenPrompt/qwen2-1.5b-instruct"),260    ],261    # Beautiful Prompt262    "BeautifulPrompt": [263        ("alibaba-pai/pai-bloom-1b1-text2prompt-sd", "config.json", "models/BeautifulPrompt/pai-bloom-1b1-text2prompt-sd"),264        ("alibaba-pai/pai-bloom-1b1-text2prompt-sd", "generation_config.json", "models/BeautifulPrompt/pai-bloom-1b1-text2prompt-sd"),265        ("alibaba-pai/pai-bloom-1b1-text2prompt-sd", "model.safetensors", "models/BeautifulPrompt/pai-bloom-1b1-text2prompt-sd"),266        ("alibaba-pai/pai-bloom-1b1-text2prompt-sd", "special_tokens_map.json", "models/BeautifulPrompt/pai-bloom-1b1-text2prompt-sd"),267        ("alibaba-pai/pai-bloom-1b1-text2prompt-sd", "tokenizer.json", "models/BeautifulPrompt/pai-bloom-1b1-text2prompt-sd"),268        ("alibaba-pai/pai-bloom-1b1-text2prompt-sd", "tokenizer_config.json", "models/BeautifulPrompt/pai-bloom-1b1-text2prompt-sd"),269    ],270    # Omost prompt271    "OmostPrompt":[272        ("lllyasviel/omost-llama-3-8b-4bits", "model-00001-of-00002.safetensors", "models/OmostPrompt/omost-llama-3-8b-4bits"),273        ("lllyasviel/omost-llama-3-8b-4bits", "model-00002-of-00002.safetensors", "models/OmostPrompt/omost-llama-3-8b-4bits"),274        ("lllyasviel/omost-llama-3-8b-4bits", "tokenizer.json", "models/OmostPrompt/omost-llama-3-8b-4bits"),275        ("lllyasviel/omost-llama-3-8b-4bits", "tokenizer_config.json", "models/OmostPrompt/omost-llama-3-8b-4bits"),  276        ("lllyasviel/omost-llama-3-8b-4bits", "config.json", "models/OmostPrompt/omost-llama-3-8b-4bits"),277        ("lllyasviel/omost-llama-3-8b-4bits", "generation_config.json", "models/OmostPrompt/omost-llama-3-8b-4bits"),278        ("lllyasviel/omost-llama-3-8b-4bits", "model.safetensors.index.json", "models/OmostPrompt/omost-llama-3-8b-4bits"),279        ("lllyasviel/omost-llama-3-8b-4bits", "special_tokens_map.json", "models/OmostPrompt/omost-llama-3-8b-4bits"),280    ],281    # Translator282    "opus-mt-zh-en": [283        ("Helsinki-NLP/opus-mt-zh-en", "config.json", "models/translator/opus-mt-zh-en"),284        ("Helsinki-NLP/opus-mt-zh-en", "generation_config.json", "models/translator/opus-mt-zh-en"),285        ("Helsinki-NLP/opus-mt-zh-en", "metadata.json", "models/translator/opus-mt-zh-en"),286        ("Helsinki-NLP/opus-mt-zh-en", "pytorch_model.bin", "models/translator/opus-mt-zh-en"),287        ("Helsinki-NLP/opus-mt-zh-en", "source.spm", "models/translator/opus-mt-zh-en"),288        ("Helsinki-NLP/opus-mt-zh-en", "target.spm", "models/translator/opus-mt-zh-en"),289        ("Helsinki-NLP/opus-mt-zh-en", "tokenizer_config.json", "models/translator/opus-mt-zh-en"),290        ("Helsinki-NLP/opus-mt-zh-en", "vocab.json", "models/translator/opus-mt-zh-en"),291    ],292    # IP-Adapter293    "IP-Adapter-SD": [294        ("h94/IP-Adapter", "models/image_encoder/model.safetensors", "models/IpAdapter/stable_diffusion/image_encoder"),295        ("h94/IP-Adapter", "models/ip-adapter_sd15.bin", "models/IpAdapter/stable_diffusion"),296    ],297    "IP-Adapter-SDXL": [298        ("h94/IP-Adapter", "sdxl_models/image_encoder/model.safetensors", "models/IpAdapter/stable_diffusion_xl/image_encoder"),299        ("h94/IP-Adapter", "sdxl_models/ip-adapter_sdxl.bin", "models/IpAdapter/stable_diffusion_xl"),300    ],301    "SDXL-vae-fp16-fix": [302        ("madebyollin/sdxl-vae-fp16-fix", "diffusion_pytorch_model.safetensors", "models/sdxl-vae-fp16-fix")303    ],304    # Kolors305    "Kolors": [306        ("Kwai-Kolors/Kolors", "text_encoder/config.json", "models/kolors/Kolors/text_encoder"),307        ("Kwai-Kolors/Kolors", "text_encoder/pytorch_model.bin.index.json", "models/kolors/Kolors/text_encoder"),308        ("Kwai-Kolors/Kolors", "text_encoder/pytorch_model-00001-of-00007.bin", "models/kolors/Kolors/text_encoder"),309        ("Kwai-Kolors/Kolors", "text_encoder/pytorch_model-00002-of-00007.bin", "models/kolors/Kolors/text_encoder"),310        ("Kwai-Kolors/Kolors", "text_encoder/pytorch_model-00003-of-00007.bin", "models/kolors/Kolors/text_encoder"),311        ("Kwai-Kolors/Kolors", "text_encoder/pytorch_model-00004-of-00007.bin", "models/kolors/Kolors/text_encoder"),312        ("Kwai-Kolors/Kolors", "text_encoder/pytorch_model-00005-of-00007.bin", "models/kolors/Kolors/text_encoder"),313        ("Kwai-Kolors/Kolors", "text_encoder/pytorch_model-00006-of-00007.bin", "models/kolors/Kolors/text_encoder"),314        ("Kwai-Kolors/Kolors", "text_encoder/pytorch_model-00007-of-00007.bin", "models/kolors/Kolors/text_encoder"),315        ("Kwai-Kolors/Kolors", "unet/diffusion_pytorch_model.safetensors", "models/kolors/Kolors/unet"),316        ("Kwai-Kolors/Kolors", "vae/diffusion_pytorch_model.safetensors", "models/kolors/Kolors/vae"),317    ],318    # FLUX319    "FLUX.1-dev": [320        ("black-forest-labs/FLUX.1-dev", "text_encoder/model.safetensors", "models/FLUX/FLUX.1-dev/text_encoder"),321        ("black-forest-labs/FLUX.1-dev", "text_encoder_2/config.json", "models/FLUX/FLUX.1-dev/text_encoder_2"),322        ("black-forest-labs/FLUX.1-dev", "text_encoder_2/model-00001-of-00002.safetensors", "models/FLUX/FLUX.1-dev/text_encoder_2"),323        ("black-forest-labs/FLUX.1-dev", "text_encoder_2/model-00002-of-00002.safetensors", "models/FLUX/FLUX.1-dev/text_encoder_2"),324        ("black-forest-labs/FLUX.1-dev", "text_encoder_2/model.safetensors.index.json", "models/FLUX/FLUX.1-dev/text_encoder_2"),325        ("black-forest-labs/FLUX.1-dev", "ae.safetensors", "models/FLUX/FLUX.1-dev"),326        ("black-forest-labs/FLUX.1-dev", "flux1-dev.safetensors", "models/FLUX/FLUX.1-dev"),327    ],328    "InstantX/FLUX.1-dev-IP-Adapter": {329        "file_list": [330            ("InstantX/FLUX.1-dev-IP-Adapter", "ip-adapter.bin", "models/IpAdapter/InstantX/FLUX.1-dev-IP-Adapter"),331            ("google/siglip-so400m-patch14-384", "model.safetensors", "models/IpAdapter/InstantX/FLUX.1-dev-IP-Adapter/image_encoder"),332            ("google/siglip-so400m-patch14-384", "config.json", "models/IpAdapter/InstantX/FLUX.1-dev-IP-Adapter/image_encoder"),333        ],334        "load_path": [335            "models/IpAdapter/InstantX/FLUX.1-dev-IP-Adapter/ip-adapter.bin",336            "models/IpAdapter/InstantX/FLUX.1-dev-IP-Adapter/image_encoder",337        ],338    },339    # RIFE340    "RIFE": [341        ("AlexWortega/RIFE", "flownet.pkl", "models/RIFE"),342    ],343    # CogVideo344    "CogVideoX-5B": [345        ("THUDM/CogVideoX-5b", "text_encoder/config.json", "models/CogVideo/CogVideoX-5b/text_encoder"),346        ("THUDM/CogVideoX-5b", "text_encoder/model.safetensors.index.json", "models/CogVideo/CogVideoX-5b/text_encoder"),347        ("THUDM/CogVideoX-5b", "text_encoder/model-00001-of-00002.safetensors", "models/CogVideo/CogVideoX-5b/text_encoder"),348        ("THUDM/CogVideoX-5b", "text_encoder/model-00002-of-00002.safetensors", "models/CogVideo/CogVideoX-5b/text_encoder"),349        ("THUDM/CogVideoX-5b", "transformer/config.json", "models/CogVideo/CogVideoX-5b/transformer"),350        ("THUDM/CogVideoX-5b", "transformer/diffusion_pytorch_model.safetensors.index.json", "models/CogVideo/CogVideoX-5b/transformer"),351        ("THUDM/CogVideoX-5b", "transformer/diffusion_pytorch_model-00001-of-00002.safetensors", "models/CogVideo/CogVideoX-5b/transformer"),352        ("THUDM/CogVideoX-5b", "transformer/diffusion_pytorch_model-00002-of-00002.safetensors", "models/CogVideo/CogVideoX-5b/transformer"),353        ("THUDM/CogVideoX-5b", "vae/diffusion_pytorch_model.safetensors", "models/CogVideo/CogVideoX-5b/vae"),354    ],355    # Stable Diffusion 3.5356    "StableDiffusion3.5-large": [357        ("stabilityai/stable-diffusion-3.5-large", "sd3.5_large.safetensors", "models/stable_diffusion_3"),358        ("stabilityai/stable-diffusion-3.5-large", "text_encoders/clip_l.safetensors", "models/stable_diffusion_3/text_encoders"),359        ("stabilityai/stable-diffusion-3.5-large", "text_encoders/clip_g.safetensors", "models/stable_diffusion_3/text_encoders"),360        ("stabilityai/stable-diffusion-3.5-large", "text_encoders/t5xxl_fp16.safetensors", "models/stable_diffusion_3/text_encoders"),361    ],362}363preset_models_on_modelscope = {364    # Hunyuan DiT365    "HunyuanDiT": [366        ("modelscope/HunyuanDiT", "t2i/clip_text_encoder/pytorch_model.bin", "models/HunyuanDiT/t2i/clip_text_encoder"),367        ("modelscope/HunyuanDiT", "t2i/mt5/pytorch_model.bin", "models/HunyuanDiT/t2i/mt5"),368        ("modelscope/HunyuanDiT", "t2i/model/pytorch_model_ema.pt", "models/HunyuanDiT/t2i/model"),369        ("modelscope/HunyuanDiT", "t2i/sdxl-vae-fp16-fix/diffusion_pytorch_model.bin", "models/HunyuanDiT/t2i/sdxl-vae-fp16-fix"),370    ],371    # Stable Video Diffusion372    "stable-video-diffusion-img2vid-xt": [373        ("AI-ModelScope/stable-video-diffusion-img2vid-xt", "svd_xt.safetensors", "models/stable_video_diffusion"),374    ],375    # ExVideo376    "ExVideo-SVD-128f-v1": [377        ("ECNU-CILab/ExVideo-SVD-128f-v1", "model.fp16.safetensors", "models/stable_video_diffusion"),378    ],379    "ExVideo-CogVideoX-LoRA-129f-v1": [380        ("ECNU-CILab/ExVideo-CogVideoX-LoRA-129f-v1", "ExVideo-CogVideoX-LoRA-129f-v1.safetensors", "models/lora"),381    ],382    # Stable Diffusion383    "StableDiffusion_v15": [384        ("AI-ModelScope/stable-diffusion-v1-5", "v1-5-pruned-emaonly.safetensors", "models/stable_diffusion"),385    ],386    "DreamShaper_8": [387        ("sd_lora/dreamshaper_8", "dreamshaper_8.safetensors", "models/stable_diffusion"),388    ],389    "AingDiffusion_v12": [390        ("sd_lora/aingdiffusion_v12", "aingdiffusion_v12.safetensors", "models/stable_diffusion"),391    ],392    "Flat2DAnimerge_v45Sharp": [393        ("sd_lora/Flat-2D-Animerge", "flat2DAnimerge_v45Sharp.safetensors", "models/stable_diffusion"),394    ],395    # Textual Inversion396    "TextualInversion_VeryBadImageNegative_v1.3": [397        ("sd_lora/verybadimagenegative_v1.3", "verybadimagenegative_v1.3.pt", "models/textual_inversion"),398    ],399    # Stable Diffusion XL400    "StableDiffusionXL_v1": [401        ("AI-ModelScope/stable-diffusion-xl-base-1.0", "sd_xl_base_1.0.safetensors", "models/stable_diffusion_xl"),402    ],403    "BluePencilXL_v200": [404        ("sd_lora/bluePencilXL_v200", "bluePencilXL_v200.safetensors", "models/stable_diffusion_xl"),405    ],406    "StableDiffusionXL_Turbo": [407        ("AI-ModelScope/sdxl-turbo", "sd_xl_turbo_1.0_fp16.safetensors", "models/stable_diffusion_xl_turbo"),408    ],409    "SDXL_lora_zyd232_ChineseInkStyle_SDXL_v1_0": [410        ("sd_lora/zyd232_ChineseInkStyle_SDXL_v1_0", "zyd232_ChineseInkStyle_SDXL_v1_0.safetensors", "models/lora"),411    ],412    # Stable Diffusion 3413    "StableDiffusion3": [414        ("AI-ModelScope/stable-diffusion-3-medium", "sd3_medium_incl_clips_t5xxlfp16.safetensors", "models/stable_diffusion_3"),415    ],416    "StableDiffusion3_without_T5": [417        ("AI-ModelScope/stable-diffusion-3-medium", "sd3_medium_incl_clips.safetensors", "models/stable_diffusion_3"),418    ],419    # ControlNet420    "ControlNet_v11f1p_sd15_depth": [421        ("AI-ModelScope/ControlNet-v1-1", "control_v11f1p_sd15_depth.pth", "models/ControlNet"),422        ("sd_lora/Annotators", "dpt_hybrid-midas-501f0c75.pt", "models/Annotators")423    ],424    "ControlNet_v11p_sd15_softedge": [425        ("AI-ModelScope/ControlNet-v1-1", "control_v11p_sd15_softedge.pth", "models/ControlNet"),426        ("sd_lora/Annotators", "ControlNetHED.pth", "models/Annotators")427    ],428    "ControlNet_v11f1e_sd15_tile": [429        ("AI-ModelScope/ControlNet-v1-1", "control_v11f1e_sd15_tile.pth", "models/ControlNet")430    ],431    "ControlNet_v11p_sd15_lineart": [432        ("AI-ModelScope/ControlNet-v1-1", "control_v11p_sd15_lineart.pth", "models/ControlNet"),433        ("sd_lora/Annotators", "sk_model.pth", "models/Annotators"),434        ("sd_lora/Annotators", "sk_model2.pth", "models/Annotators")435    ],436    "ControlNet_union_sdxl_promax": [437        ("AI-ModelScope/controlnet-union-sdxl-1.0", "diffusion_pytorch_model_promax.safetensors", "models/ControlNet/controlnet_union"),438        ("sd_lora/Annotators", "dpt_hybrid-midas-501f0c75.pt", "models/Annotators")439    ],440    "Annotators:Depth": [441        ("sd_lora/Annotators", "dpt_hybrid-midas-501f0c75.pt", "models/Annotators"),442    ],443    "Annotators:Softedge": [444        ("sd_lora/Annotators", "ControlNetHED.pth", "models/Annotators"),445    ],446    "Annotators:Lineart": [447        ("sd_lora/Annotators", "sk_model.pth", "models/Annotators"),448        ("sd_lora/Annotators", "sk_model2.pth", "models/Annotators"),449    ],450    "Annotators:Normal": [451        ("sd_lora/Annotators", "scannet.pt", "models/Annotators"),452    ],453    "Annotators:Openpose": [454        ("sd_lora/Annotators", "body_pose_model.pth", "models/Annotators"),455        ("sd_lora/Annotators", "facenet.pth", "models/Annotators"),456        ("sd_lora/Annotators", "hand_pose_model.pth", "models/Annotators"),457    ],458    # AnimateDiff459    "AnimateDiff_v2": [460        ("Shanghai_AI_Laboratory/animatediff", "mm_sd_v15_v2.ckpt", "models/AnimateDiff"),461    ],462    "AnimateDiff_xl_beta": [463        ("Shanghai_AI_Laboratory/animatediff", "mm_sdxl_v10_beta.ckpt", "models/AnimateDiff"),464    ],465    # RIFE466    "RIFE": [467        ("Damo_XR_Lab/cv_rife_video-frame-interpolation", "flownet.pkl", "models/RIFE"),468    ],469    # Qwen Prompt470    "QwenPrompt": {471        "file_list": [472            ("qwen/Qwen2-1.5B-Instruct", "config.json", "models/QwenPrompt/qwen2-1.5b-instruct"),473            ("qwen/Qwen2-1.5B-Instruct", "generation_config.json", "models/QwenPrompt/qwen2-1.5b-instruct"),474            ("qwen/Qwen2-1.5B-Instruct", "model.safetensors", "models/QwenPrompt/qwen2-1.5b-instruct"),475            ("qwen/Qwen2-1.5B-Instruct", "special_tokens_map.json", "models/QwenPrompt/qwen2-1.5b-instruct"),476            ("qwen/Qwen2-1.5B-Instruct", "tokenizer.json", "models/QwenPrompt/qwen2-1.5b-instruct"),477            ("qwen/Qwen2-1.5B-Instruct", "tokenizer_config.json", "models/QwenPrompt/qwen2-1.5b-instruct"),478            ("qwen/Qwen2-1.5B-Instruct", "merges.txt", "models/QwenPrompt/qwen2-1.5b-instruct"),479            ("qwen/Qwen2-1.5B-Instruct", "vocab.json", "models/QwenPrompt/qwen2-1.5b-instruct"),480        ],481        "load_path": [482            "models/QwenPrompt/qwen2-1.5b-instruct",483        ],484    },485    # Beautiful Prompt486    "BeautifulPrompt": {487        "file_list": [488            ("AI-ModelScope/pai-bloom-1b1-text2prompt-sd", "config.json", "models/BeautifulPrompt/pai-bloom-1b1-text2prompt-sd"),489            ("AI-ModelScope/pai-bloom-1b1-text2prompt-sd", "generation_config.json", "models/BeautifulPrompt/pai-bloom-1b1-text2prompt-sd"),490            ("AI-ModelScope/pai-bloom-1b1-text2prompt-sd", "model.safetensors", "models/BeautifulPrompt/pai-bloom-1b1-text2prompt-sd"),491            ("AI-ModelScope/pai-bloom-1b1-text2prompt-sd", "special_tokens_map.json", "models/BeautifulPrompt/pai-bloom-1b1-text2prompt-sd"),492            ("AI-ModelScope/pai-bloom-1b1-text2prompt-sd", "tokenizer.json", "models/BeautifulPrompt/pai-bloom-1b1-text2prompt-sd"),493            ("AI-ModelScope/pai-bloom-1b1-text2prompt-sd", "tokenizer_config.json", "models/BeautifulPrompt/pai-bloom-1b1-text2prompt-sd"),494        ],495        "load_path": [496            "models/BeautifulPrompt/pai-bloom-1b1-text2prompt-sd",497        ],498    },499    # Omost prompt500    "OmostPrompt": {501        "file_list": [502            ("Omost/omost-llama-3-8b-4bits", "model-00001-of-00002.safetensors", "models/OmostPrompt/omost-llama-3-8b-4bits"),503            ("Omost/omost-llama-3-8b-4bits", "model-00002-of-00002.safetensors", "models/OmostPrompt/omost-llama-3-8b-4bits"),504            ("Omost/omost-llama-3-8b-4bits", "tokenizer.json", "models/OmostPrompt/omost-llama-3-8b-4bits"),505            ("Omost/omost-llama-3-8b-4bits", "tokenizer_config.json", "models/OmostPrompt/omost-llama-3-8b-4bits"),  506            ("Omost/omost-llama-3-8b-4bits", "config.json", "models/OmostPrompt/omost-llama-3-8b-4bits"),507            ("Omost/omost-llama-3-8b-4bits", "generation_config.json", "models/OmostPrompt/omost-llama-3-8b-4bits"),508            ("Omost/omost-llama-3-8b-4bits", "model.safetensors.index.json", "models/OmostPrompt/omost-llama-3-8b-4bits"),509            ("Omost/omost-llama-3-8b-4bits", "special_tokens_map.json", "models/OmostPrompt/omost-llama-3-8b-4bits"),510        ],511        "load_path": [512            "models/OmostPrompt/omost-llama-3-8b-4bits",513        ],514    },515    # Translator516    "opus-mt-zh-en": {517        "file_list": [518            ("moxying/opus-mt-zh-en", "config.json", "models/translator/opus-mt-zh-en"),519            ("moxying/opus-mt-zh-en", "generation_config.json", "models/translator/opus-mt-zh-en"),520            ("moxying/opus-mt-zh-en", "metadata.json", "models/translator/opus-mt-zh-en"),521            ("moxying/opus-mt-zh-en", "pytorch_model.bin", "models/translator/opus-mt-zh-en"),522            ("moxying/opus-mt-zh-en", "source.spm", "models/translator/opus-mt-zh-en"),523            ("moxying/opus-mt-zh-en", "target.spm", "models/translator/opus-mt-zh-en"),524            ("moxying/opus-mt-zh-en", "tokenizer_config.json", "models/translator/opus-mt-zh-en"),525            ("moxying/opus-mt-zh-en", "vocab.json", "models/translator/opus-mt-zh-en"),526        ],527        "load_path": [528            "models/translator/opus-mt-zh-en",529        ],530    },531    # IP-Adapter532    "IP-Adapter-SD": [533        ("AI-ModelScope/IP-Adapter", "models/image_encoder/model.safetensors", "models/IpAdapter/stable_diffusion/image_encoder"),534        ("AI-ModelScope/IP-Adapter", "models/ip-adapter_sd15.bin", "models/IpAdapter/stable_diffusion"),535    ],536    "IP-Adapter-SDXL": [537        ("AI-ModelScope/IP-Adapter", "sdxl_models/image_encoder/model.safetensors", "models/IpAdapter/stable_diffusion_xl/image_encoder"),538        ("AI-ModelScope/IP-Adapter", "sdxl_models/ip-adapter_sdxl.bin", "models/IpAdapter/stable_diffusion_xl"),539    ],540    # Kolors541    "Kolors": {542        "file_list": [543            ("Kwai-Kolors/Kolors", "text_encoder/config.json", "models/kolors/Kolors/text_encoder"),544            ("Kwai-Kolors/Kolors", "text_encoder/pytorch_model.bin.index.json", "models/kolors/Kolors/text_encoder"),545            ("Kwai-Kolors/Kolors", "text_encoder/pytorch_model-00001-of-00007.bin", "models/kolors/Kolors/text_encoder"),546            ("Kwai-Kolors/Kolors", "text_encoder/pytorch_model-00002-of-00007.bin", "models/kolors/Kolors/text_encoder"),547            ("Kwai-Kolors/Kolors", "text_encoder/pytorch_model-00003-of-00007.bin", "models/kolors/Kolors/text_encoder"),548            ("Kwai-Kolors/Kolors", "text_encoder/pytorch_model-00004-of-00007.bin", "models/kolors/Kolors/text_encoder"),549            ("Kwai-Kolors/Kolors", "text_encoder/pytorch_model-00005-of-00007.bin", "models/kolors/Kolors/text_encoder"),550            ("Kwai-Kolors/Kolors", "text_encoder/pytorch_model-00006-of-00007.bin", "models/kolors/Kolors/text_encoder"),551            ("Kwai-Kolors/Kolors", "text_encoder/pytorch_model-00007-of-00007.bin", "models/kolors/Kolors/text_encoder"),552            ("Kwai-Kolors/Kolors", "unet/diffusion_pytorch_model.safetensors", "models/kolors/Kolors/unet"),553            ("Kwai-Kolors/Kolors", "vae/diffusion_pytorch_model.safetensors", "models/kolors/Kolors/vae"),554        ],555        "load_path": [556            "models/kolors/Kolors/text_encoder",557            "models/kolors/Kolors/unet/diffusion_pytorch_model.safetensors",558            "models/kolors/Kolors/vae/diffusion_pytorch_model.safetensors",559        ],560    },561    "SDXL-vae-fp16-fix": [562        ("AI-ModelScope/sdxl-vae-fp16-fix", "diffusion_pytorch_model.safetensors", "models/sdxl-vae-fp16-fix")563    ],564    # FLUX565    "FLUX.1-dev": {566        "file_list": [567            ("AI-ModelScope/FLUX.1-dev", "text_encoder/model.safetensors", "models/FLUX/FLUX.1-dev/text_encoder"),568            ("AI-ModelScope/FLUX.1-dev", "text_encoder_2/config.json", "models/FLUX/FLUX.1-dev/text_encoder_2"),569            ("AI-ModelScope/FLUX.1-dev", "text_encoder_2/model-00001-of-00002.safetensors", "models/FLUX/FLUX.1-dev/text_encoder_2"),570            ("AI-ModelScope/FLUX.1-dev", "text_encoder_2/model-00002-of-00002.safetensors", "models/FLUX/FLUX.1-dev/text_encoder_2"),571            ("AI-ModelScope/FLUX.1-dev", "text_encoder_2/model.safetensors.index.json", "models/FLUX/FLUX.1-dev/text_encoder_2"),572            ("AI-ModelScope/FLUX.1-dev", "ae.safetensors", "models/FLUX/FLUX.1-dev"),573            ("AI-ModelScope/FLUX.1-dev", "flux1-dev.safetensors", "models/FLUX/FLUX.1-dev"),574        ],575        "load_path": [576            "models/FLUX/FLUX.1-dev/text_encoder/model.safetensors",577            "models/FLUX/FLUX.1-dev/text_encoder_2",578            "models/FLUX/FLUX.1-dev/ae.safetensors",579            "models/FLUX/FLUX.1-dev/flux1-dev.safetensors"580        ],581    },582    "FLUX.1-schnell": {583        "file_list": [584            ("AI-ModelScope/FLUX.1-dev", "text_encoder/model.safetensors", "models/FLUX/FLUX.1-dev/text_encoder"),585            ("AI-ModelScope/FLUX.1-dev", "text_encoder_2/config.json", "models/FLUX/FLUX.1-dev/text_encoder_2"),586            ("AI-ModelScope/FLUX.1-dev", "text_encoder_2/model-00001-of-00002.safetensors", "models/FLUX/FLUX.1-dev/text_encoder_2"),587            ("AI-ModelScope/FLUX.1-dev", "text_encoder_2/model-00002-of-00002.safetensors", "models/FLUX/FLUX.1-dev/text_encoder_2"),588            ("AI-ModelScope/FLUX.1-dev", "text_encoder_2/model.safetensors.index.json", "models/FLUX/FLUX.1-dev/text_encoder_2"),589            ("AI-ModelScope/FLUX.1-dev", "ae.safetensors", "models/FLUX/FLUX.1-dev"),590            ("AI-ModelScope/FLUX.1-schnell", "flux1-schnell.safetensors", "models/FLUX/FLUX.1-schnell"),591        ],592        "load_path": [593            "models/FLUX/FLUX.1-dev/text_encoder/model.safetensors",594            "models/FLUX/FLUX.1-dev/text_encoder_2",595            "models/FLUX/FLUX.1-dev/ae.safetensors",596            "models/FLUX/FLUX.1-schnell/flux1-schnell.safetensors"597        ],598    },599    "InstantX/FLUX.1-dev-Controlnet-Union-alpha": [600        ("InstantX/FLUX.1-dev-Controlnet-Union-alpha", "diffusion_pytorch_model.safetensors", "models/ControlNet/InstantX/FLUX.1-dev-Controlnet-Union-alpha"),601    ],602    "jasperai/Flux.1-dev-Controlnet-Depth": [603        ("jasperai/Flux.1-dev-Controlnet-Depth", "diffusion_pytorch_model.safetensors", "models/ControlNet/jasperai/Flux.1-dev-Controlnet-Depth"),604    ],605    "jasperai/Flux.1-dev-Controlnet-Surface-Normals": [606        ("jasperai/Flux.1-dev-Controlnet-Surface-Normals", "diffusion_pytorch_model.safetensors", "models/ControlNet/jasperai/Flux.1-dev-Controlnet-Surface-Normals"),607    ],608    "jasperai/Flux.1-dev-Controlnet-Upscaler": [609        ("jasperai/Flux.1-dev-Controlnet-Upscaler", "diffusion_pytorch_model.safetensors", "models/ControlNet/jasperai/Flux.1-dev-Controlnet-Upscaler"),610    ],611    "alimama-creative/FLUX.1-dev-Controlnet-Inpainting-Alpha": [612        ("alimama-creative/FLUX.1-dev-Controlnet-Inpainting-Alpha", "diffusion_pytorch_model.safetensors", "models/ControlNet/alimama-creative/FLUX.1-dev-Controlnet-Inpainting-Alpha"),613    ],614    "alimama-creative/FLUX.1-dev-Controlnet-Inpainting-Beta": [615        ("alimama-creative/FLUX.1-dev-Controlnet-Inpainting-Beta", "diffusion_pytorch_model.safetensors", "models/ControlNet/alimama-creative/FLUX.1-dev-Controlnet-Inpainting-Beta"),616    ],617    "Shakker-Labs/FLUX.1-dev-ControlNet-Depth": [618        ("Shakker-Labs/FLUX.1-dev-ControlNet-Depth", "diffusion_pytorch_model.safetensors", "models/ControlNet/Shakker-Labs/FLUX.1-dev-ControlNet-Depth"),619    ],620    "Shakker-Labs/FLUX.1-dev-ControlNet-Union-Pro": [621        ("Shakker-Labs/FLUX.1-dev-ControlNet-Union-Pro", "diffusion_pytorch_model.safetensors", "models/ControlNet/Shakker-Labs/FLUX.1-dev-ControlNet-Union-Pro"),622    ],623    "InstantX/FLUX.1-dev-IP-Adapter": {624        "file_list": [625            ("InstantX/FLUX.1-dev-IP-Adapter", "ip-adapter.bin", "models/IpAdapter/InstantX/FLUX.1-dev-IP-Adapter"),626            ("AI-ModelScope/siglip-so400m-patch14-384", "model.safetensors", "models/IpAdapter/InstantX/FLUX.1-dev-IP-Adapter/image_encoder"),627            ("AI-ModelScope/siglip-so400m-patch14-384", "config.json", "models/IpAdapter/InstantX/FLUX.1-dev-IP-Adapter/image_encoder"),628        ],629        "load_path": [630            "models/IpAdapter/InstantX/FLUX.1-dev-IP-Adapter/ip-adapter.bin",631            "models/IpAdapter/InstantX/FLUX.1-dev-IP-Adapter/image_encoder",632        ],633    },634    "InfiniteYou":{635        "file_list":[636            ("ByteDance/InfiniteYou", "infu_flux_v1.0/aes_stage2/InfuseNetModel/diffusion_pytorch_model-00001-of-00002.safetensors", "models/InfiniteYou/InfuseNetModel"),637            ("ByteDance/InfiniteYou", "infu_flux_v1.0/aes_stage2/InfuseNetModel/diffusion_pytorch_model-00002-of-00002.safetensors", "models/InfiniteYou/InfuseNetModel"),638            ("ByteDance/InfiniteYou", "infu_flux_v1.0/aes_stage2/image_proj_model.bin", "models/InfiniteYou"),639            ("ByteDance/InfiniteYou", "supports/insightface/models/antelopev2/1k3d68.onnx", "models/InfiniteYou/insightface/models/antelopev2"),640            ("ByteDance/InfiniteYou", "supports/insightface/models/antelopev2/2d106det.onnx", "models/InfiniteYou/insightface/models/antelopev2"),641            ("ByteDance/InfiniteYou", "supports/insightface/models/antelopev2/genderage.onnx", "models/InfiniteYou/insightface/models/antelopev2"),642            ("ByteDance/InfiniteYou", "supports/insightface/models/antelopev2/glintr100.onnx", "models/InfiniteYou/insightface/models/antelopev2"),643            ("ByteDance/InfiniteYou", "supports/insightface/models/antelopev2/scrfd_10g_bnkps.onnx", "models/InfiniteYou/insightface/models/antelopev2"),            644        ],645        "load_path":[646            [647                "models/InfiniteYou/InfuseNetModel/diffusion_pytorch_model-00001-of-00002.safetensors",648                "models/InfiniteYou/InfuseNetModel/diffusion_pytorch_model-00002-of-00002.safetensors"649            ],650            "models/InfiniteYou/image_proj_model.bin",651            ],652    },653    # ESRGAN654    "ESRGAN_x4": [655        ("AI-ModelScope/Real-ESRGAN", "RealESRGAN_x4.pth", "models/ESRGAN"),656    ],657    # RIFE658    "RIFE": [659        ("AI-ModelScope/RIFE", "flownet.pkl", "models/RIFE"),660    ],661    # Omnigen662    "OmniGen-v1": {663        "file_list": [664            ("BAAI/OmniGen-v1", "vae/diffusion_pytorch_model.safetensors", "models/OmniGen/OmniGen-v1/vae"),665            ("BAAI/OmniGen-v1", "model.safetensors", "models/OmniGen/OmniGen-v1"),666            ("BAAI/OmniGen-v1", "config.json", "models/OmniGen/OmniGen-v1"),667            ("BAAI/OmniGen-v1", "special_tokens_map.json", "models/OmniGen/OmniGen-v1"),668            ("BAAI/OmniGen-v1", "tokenizer_config.json", "models/OmniGen/OmniGen-v1"),669            ("BAAI/OmniGen-v1", "tokenizer.json", "models/OmniGen/OmniGen-v1"),670        ],671        "load_path": [672            "models/OmniGen/OmniGen-v1/vae/diffusion_pytorch_model.safetensors",673            "models/OmniGen/OmniGen-v1/model.safetensors",674        ]675    },676    # CogVideo677    "CogVideoX-5B": {678        "file_list": [679            ("ZhipuAI/CogVideoX-5b", "text_encoder/config.json", "models/CogVideo/CogVideoX-5b/text_encoder"),680            ("ZhipuAI/CogVideoX-5b", "text_encoder/model.safetensors.index.json", "models/CogVideo/CogVideoX-5b/text_encoder"),681            ("ZhipuAI/CogVideoX-5b", "text_encoder/model-00001-of-00002.safetensors", "models/CogVideo/CogVideoX-5b/text_encoder"),682            ("ZhipuAI/CogVideoX-5b", "text_encoder/model-00002-of-00002.safetensors", "models/CogVideo/CogVideoX-5b/text_encoder"),683            ("ZhipuAI/CogVideoX-5b", "transformer/config.json", "models/CogVideo/CogVideoX-5b/transformer"),684            ("ZhipuAI/CogVideoX-5b", "transformer/diffusion_pytorch_model.safetensors.index.json", "models/CogVideo/CogVideoX-5b/transformer"),685            ("ZhipuAI/CogVideoX-5b", "transformer/diffusion_pytorch_model-00001-of-00002.safetensors", "models/CogVideo/CogVideoX-5b/transformer"),686            ("ZhipuAI/CogVideoX-5b", "transformer/diffusion_pytorch_model-00002-of-00002.safetensors", "models/CogVideo/CogVideoX-5b/transformer"),687            ("ZhipuAI/CogVideoX-5b", "vae/diffusion_pytorch_model.safetensors", "models/CogVideo/CogVideoX-5b/vae"),688        ],689        "load_path": [690            "models/CogVideo/CogVideoX-5b/text_encoder",691            "models/CogVideo/CogVideoX-5b/transformer",692            "models/CogVideo/CogVideoX-5b/vae/diffusion_pytorch_model.safetensors",693        ],694    },695    # Stable Diffusion 3.5696    "StableDiffusion3.5-large": [697        ("AI-ModelScope/stable-diffusion-3.5-large", "sd3.5_large.safetensors", "models/stable_diffusion_3"),698        ("AI-ModelScope/stable-diffusion-3.5-large", "text_encoders/clip_l.safetensors", "models/stable_diffusion_3/text_encoders"),699        ("AI-ModelScope/stable-diffusion-3.5-large", "text_encoders/clip_g.safetensors", "models/stable_diffusion_3/text_encoders"),700        ("AI-ModelScope/stable-diffusion-3.5-large", "text_encoders/t5xxl_fp16.safetensors", "models/stable_diffusion_3/text_encoders"),701    ],702    "StableDiffusion3.5-medium": [703        ("AI-ModelScope/stable-diffusion-3.5-medium", "sd3.5_medium.safetensors", "models/stable_diffusion_3"),704        ("AI-ModelScope/stable-diffusion-3.5-large", "text_encoders/clip_l.safetensors", "models/stable_diffusion_3/text_encoders"),705        ("AI-ModelScope/stable-diffusion-3.5-large", "text_encoders/clip_g.safetensors", "models/stable_diffusion_3/text_encoders"),706        ("AI-ModelScope/stable-diffusion-3.5-large", "text_encoders/t5xxl_fp16.safetensors", "models/stable_diffusion_3/text_encoders"),707    ],708    "StableDiffusion3.5-large-turbo": [709        ("AI-ModelScope/stable-diffusion-3.5-large-turbo", "sd3.5_large_turbo.safetensors", "models/stable_diffusion_3"),710        ("AI-ModelScope/stable-diffusion-3.5-large", "text_encoders/clip_l.safetensors", "models/stable_diffusion_3/text_encoders"),711        ("AI-ModelScope/stable-diffusion-3.5-large", "text_encoders/clip_g.safetensors", "models/stable_diffusion_3/text_encoders"),712        ("AI-ModelScope/stable-diffusion-3.5-large", "text_encoders/t5xxl_fp16.safetensors", "models/stable_diffusion_3/text_encoders"),713    ],714    "HunyuanVideo":{715        "file_list": [716            ("AI-ModelScope/clip-vit-large-patch14", "model.safetensors", "models/HunyuanVideo/text_encoder"),717            ("DiffSynth-Studio/HunyuanVideo_MLLM_text_encoder", "model-00001-of-00004.safetensors", "models/HunyuanVideo/text_encoder_2"),718            ("DiffSynth-Studio/HunyuanVideo_MLLM_text_encoder", "model-00002-of-00004.safetensors", "models/HunyuanVideo/text_encoder_2"),719            ("DiffSynth-Studio/HunyuanVideo_MLLM_text_encoder", "model-00003-of-00004.safetensors", "models/HunyuanVideo/text_encoder_2"),720            ("DiffSynth-Studio/HunyuanVideo_MLLM_text_encoder", "model-00004-of-00004.safetensors", "models/HunyuanVideo/text_encoder_2"),721            ("DiffSynth-Studio/HunyuanVideo_MLLM_text_encoder", "config.json", "models/HunyuanVideo/text_encoder_2"),722            ("DiffSynth-Studio/HunyuanVideo_MLLM_text_encoder", "model.safetensors.index.json", "models/HunyuanVideo/text_encoder_2"),723            ("AI-ModelScope/HunyuanVideo", "hunyuan-video-t2v-720p/vae/pytorch_model.pt", "models/HunyuanVideo/vae"),724            ("AI-ModelScope/HunyuanVideo", "hunyuan-video-t2v-720p/transformers/mp_rank_00_model_states.pt", "models/HunyuanVideo/transformers")725        ],726        "load_path": [727            "models/HunyuanVideo/text_encoder/model.safetensors",728            "models/HunyuanVideo/text_encoder_2",729            "models/HunyuanVideo/vae/pytorch_model.pt",730            "models/HunyuanVideo/transformers/mp_rank_00_model_states.pt"731        ],732    },733    "HunyuanVideoI2V":{734        "file_list": [735            ("AI-ModelScope/clip-vit-large-patch14", "model.safetensors", "models/HunyuanVideoI2V/text_encoder"),736            ("AI-ModelScope/llava-llama-3-8b-v1_1-transformers", "model-00001-of-00004.safetensors", "models/HunyuanVideoI2V/text_encoder_2"),737            ("AI-ModelScope/llava-llama-3-8b-v1_1-transformers", "model-00002-of-00004.safetensors", "models/HunyuanVideoI2V/text_encoder_2"),738            ("AI-ModelScope/llava-llama-3-8b-v1_1-transformers", "model-00003-of-00004.safetensors", "models/HunyuanVideoI2V/text_encoder_2"),739            ("AI-ModelScope/llava-llama-3-8b-v1_1-transformers", "model-00004-of-00004.safetensors", "models/HunyuanVideoI2V/text_encoder_2"),740            ("AI-ModelScope/llava-llama-3-8b-v1_1-transformers", "config.json", "models/HunyuanVideoI2V/text_encoder_2"),741            ("AI-ModelScope/llava-llama-3-8b-v1_1-transformers", "model.safetensors.index.json", "models/HunyuanVideoI2V/text_encoder_2"),742            ("AI-ModelScope/HunyuanVideo-I2V", "hunyuan-video-i2v-720p/vae/pytorch_model.pt", "models/HunyuanVideoI2V/vae"),743            ("AI-ModelScope/HunyuanVideo-I2V", "hunyuan-video-i2v-720p/transformers/mp_rank_00_model_states.pt", "models/HunyuanVideoI2V/transformers")744        ],745        "load_path": [746            "models/HunyuanVideoI2V/text_encoder/model.safetensors",747            "models/HunyuanVideoI2V/text_encoder_2",748            "models/HunyuanVideoI2V/vae/pytorch_model.pt",749            "models/HunyuanVideoI2V/transformers/mp_rank_00_model_states.pt"750        ],751    },752    "HunyuanVideo-fp8":{753        "file_list": [754            ("AI-ModelScope/clip-vit-large-patch14", "model.safetensors", "models/HunyuanVideo/text_encoder"),755            ("DiffSynth-Studio/HunyuanVideo_MLLM_text_encoder", "model-00001-of-00004.safetensors", "models/HunyuanVideo/text_encoder_2"),756            ("DiffSynth-Studio/HunyuanVideo_MLLM_text_encoder", "model-00002-of-00004.safetensors", "models/HunyuanVideo/text_encoder_2"),757            ("DiffSynth-Studio/HunyuanVideo_MLLM_text_encoder", "model-00003-of-00004.safetensors", "models/HunyuanVideo/text_encoder_2"),758            ("DiffSynth-Studio/HunyuanVideo_MLLM_text_encoder", "model-00004-of-00004.safetensors", "models/HunyuanVideo/text_encoder_2"),759            ("DiffSynth-Studio/HunyuanVideo_MLLM_text_encoder", "config.json", "models/HunyuanVideo/text_encoder_2"),760            ("DiffSynth-Studio/HunyuanVideo_MLLM_text_encoder", "model.safetensors.index.json", "models/HunyuanVideo/text_encoder_2"),761            ("AI-ModelScope/HunyuanVideo", "hunyuan-video-t2v-720p/vae/pytorch_model.pt", "models/HunyuanVideo/vae"),762            ("DiffSynth-Studio/HunyuanVideo-safetensors", "model.fp8.safetensors", "models/HunyuanVideo/transformers")763        ],764        "load_path": [765            "models/HunyuanVideo/text_encoder/model.safetensors",766            "models/HunyuanVideo/text_encoder_2",767            "models/HunyuanVideo/vae/pytorch_model.pt",768            "models/HunyuanVideo/transformers/model.fp8.safetensors"769        ],770    },771}772Preset_model_id: TypeAlias = Literal[773    "HunyuanDiT",774    "stable-video-diffusion-img2vid-xt",775    "ExVideo-SVD-128f-v1",776    "ExVideo-CogVideoX-LoRA-129f-v1",777    "StableDiffusion_v15",778    "DreamShaper_8",779    "AingDiffusion_v12",780    "Flat2DAnimerge_v45Sharp",781    "TextualInversion_VeryBadImageNegative_v1.3",782    "StableDiffusionXL_v1",783    "BluePencilXL_v200",784    "StableDiffusionXL_Turbo",785    "ControlNet_v11f1p_sd15_depth",786    "ControlNet_v11p_sd15_softedge",787    "ControlNet_v11f1e_sd15_tile",788    "ControlNet_v11p_sd15_lineart",789    "AnimateDiff_v2",790    "AnimateDiff_xl_beta",791    "RIFE",792    "BeautifulPrompt",793    "opus-mt-zh-en",794    "IP-Adapter-SD",795    "IP-Adapter-SDXL",796    "StableDiffusion3",797    "StableDiffusion3_without_T5",798    "Kolors",799    "SDXL-vae-fp16-fix",800    "ControlNet_union_sdxl_promax",801    "FLUX.1-dev",802    "FLUX.1-schnell",803    "InstantX/FLUX.1-dev-Controlnet-Union-alpha",804    "jasperai/Flux.1-dev-Controlnet-Depth",805    "jasperai/Flux.1-dev-Controlnet-Surface-Normals",806    "jasperai/Flux.1-dev-Controlnet-Upscaler",807    "alimama-creative/FLUX.1-dev-Controlnet-Inpainting-Alpha",808    "alimama-creative/FLUX.1-dev-Controlnet-Inpainting-Beta",809    "Shakker-Labs/FLUX.1-dev-ControlNet-Depth",810    "Shakker-Labs/FLUX.1-dev-ControlNet-Union-Pro",811    "InstantX/FLUX.1-dev-IP-Adapter",812    "InfiniteYou",813    "SDXL_lora_zyd232_ChineseInkStyle_SDXL_v1_0",814    "QwenPrompt",815    "OmostPrompt",816    "ESRGAN_x4",817    "RIFE",818    "OmniGen-v1",819    "CogVideoX-5B",820    "Annotators:Depth",821    "Annotators:Softedge",822    "Annotators:Lineart",823    "Annotators:Normal",824    "Annotators:Openpose",825    "StableDiffusion3.5-large",826    "StableDiffusion3.5-medium",827    "HunyuanVideo",828    "HunyuanVideo-fp8",829    "HunyuanVideoI2V",830]831