+h-i SSKJrJrJrJr SSKrSSKrSSKJr SSKJ r SSK J r SSK J r JrJrJr \"5(aSSKJrJr \"5(a SS KJrJrJr \R0"\5rS rS r\S 5r"S S5rg))DictListOptionalUnionN)validate_hf_hub_args)nn)load_state_dict)_get_model_fileis_accelerate_availableis_transformers_availablelogging)PreTrainedModelPreTrainedTokenizer)AlignDevicesHook CpuOffloadremove_hook_from_modulezlearned_embeds.binzlearned_embeds.safetensorsc RURSS5nURSS5nURSS5nURSS5nURSS5nURSS5nURSS5nURS S5n URS S5n Sn U cS n S n S S S.n /n UHn[U[[R45(dSnU (aU bU bTU R S5(a>[ UU =(d [UUUUUUUU S9 n[RRUSS9nUc*[ UU =(d [UUUUUUUU S9 n[U5nOUnU RW5 M U $![anU (dUeSnSnANaSnAff=f)N cache_dirforce_downloadFproxieslocal_files_onlyhf_tokenrevision subfolder weight_nameuse_safetensorsTtext_inversionpytorch) file_type frameworkz .safetensors) weights_namerrrrtokenrr user_agentcpudevice)pop isinstancedicttorchTensorendswithr TEXT_INVERSION_NAME_SAFE safetensors load_file ExceptionTEXT_INVERSION_NAMEr append)pretrained_model_name_or_pathskwargsrrrrrrrrr allow_pickler$ state_dictspretrained_model_name_or_path model_file state_dictes ]/home/james-whalen/.local/lib/python3.13/site-packages/diffusers/loaders/textual_inversion.py"load_textual_inversion_state_dictsr=%s ;-IZZ 0%8NjjD)Gzz"4d;zz*d+Hzz*d+H ;-I**]D1Kjj!2D9OL &JK)G%7$ 9MNNJ K$7'K,@,@,P,P&!05%0%L4L"+'5 ')9&!)"+#- "J"-!2!2!>**+,0099:!=   >>**+,0099:!=  - . 2s;Y7Z^abh^i7i=cB`>a=bcBCFGMCNBOO=>  $*;6a6 ; s< !C $5 5Z[aZbcd d 6  bchbiijkl$++-- UG#tuzt{|op   e $  ! !) ,C":F))rWc /n/n[X5GH'upVUS3UR5;a^U/nSnUSU3UR;a2URUSU35 US- nUSU3UR;aM2[ SUSUS35e[ UR 5S:=(a UR SS:n U (aJX6/[SUR S5Vs/sH oSU3PM sn-- nXEV s/sHoPM sn - nMX6/- nU[ UR 5S:aUS/OU/- nGM* X44$s snfs sn f)N_1rKrLzMulti-vector Token zU already in tokenizer vocabulary. Please choose a different token name or remove the rir)rjrprPr3rZr^shaperange) rR embeddingsrBrqrrrtr#multi_vector_tokensrUis_multi_vectorr;s r<_extend_tokens_and_embeddings9TextualInversionLoaderMixin._extend_tokens_and_embeddingss  #J 7 I|y2244',g#qn (F(FF'..%!~>FAqn (F(FF!)*=)>?TUhTii\]")//2Q6Q9??1;MPQ;QOgq)//Z[J\A](^A]A7!A3A](^^^ i"8i1i"88g% C 4H14L9Q<.S\R]]%!8()) )_"8s D? ; ENr8r#r_rc :^U=(d [USS5nU=(d [USS5n[U[5(dU/OUn[U[5(dU/OUnUScU[U5-nUR X4Xg5 [ U40UD6n[U5S:a{[U5S:Xal[US[ R5(aJ[US5n[U5[U5:wa$[S[U5S[U5S35eURXxU5upyURXyU5upyUR5RRS m[U4S jU 55(a [S 5eS n S n URGcUR R#5Hup[U [$R&5(dM&[)U S 5(dM9[[U S 5[*5n [[U S 5[,5=(dD [)U R.S5=(a' [U R.R0S[,5n [2R5S5 U (d U (dM[7XS9 M UR8nUR:nUR=[U5[U5-5 UR5Rn[?Xy5HPunnURAU5 URCU5nUURDU'[2R5SUS35 MR URGXS9 U (aURIUS9 gU (aURKUS9 gg)u Load Textual Inversion embeddings into the text encoder of [`StableDiffusionPipeline`] (both 🤗 Diffusers and Automatic1111 formats are supported). Parameters: pretrained_model_name_or_path (`str` or `os.PathLike` or `List[str or os.PathLike]` or `Dict` or `List[Dict]`): Can be either one of the following or a list of them: - A string, the *model id* (for example `sd-concepts-library/low-poly-hd-logos-icons`) of a pretrained model hosted on the Hub. - A path to a *directory* (for example `./my_text_inversion_directory/`) containing the textual inversion weights. - A path to a *file* (for example `./my_text_inversions.pt`) containing textual inversion weights. - A [torch state dict](https://pytorch.org/tutorials/beginner/saving_loading_models.html#what-is-a-state-dict). token (`str` or `List[str]`, *optional*): Override the token to use for the textual inversion weights. If `pretrained_model_name_or_path` is a list, then `token` must also be a list of equal length. text_encoder ([`~transformers.CLIPTextModel`], *optional*): Frozen text-encoder ([clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14)). If not specified, function will take self.tokenizer. tokenizer ([`~transformers.CLIPTokenizer`], *optional*): A `CLIPTokenizer` to tokenize text. If not specified, function will take self.tokenizer. weight_name (`str`, *optional*): Name of a custom weight file. This should be used when: - The saved textual inversion file is in 🤗 Diffusers format, but was saved under a specific weight name such as `text_inv.bin`. - The saved textual inversion file is in the Automatic1111 format. cache_dir (`Union[str, os.PathLike]`, *optional*): Path to a directory where a downloaded pretrained model configuration is cached if the standard cache is not used. force_download (`bool`, *optional*, defaults to `False`): Whether or not to force the (re-)download of the model weights and configuration files, overriding the cached versions if they exist. proxies (`Dict[str, str]`, *optional*): A dictionary of proxy servers to use by protocol or endpoint, for example, `{'http': 'foo.bar:3128', 'http://hostname': 'foo.bar:4012'}`. The proxies are used on each request. local_files_only (`bool`, *optional*, defaults to `False`): Whether to only load local model weights and configuration files or not. If set to `True`, the model won't be downloaded from the Hub. hf_token (`str` or *bool*, *optional*): The token to use as HTTP bearer authorization for remote files. If `True`, the token generated from `diffusers-cli login` (stored in `~/.huggingface`) is used. revision (`str`, *optional*, defaults to `"main"`): The specific model version to use. It can be a branch name, a tag name, a commit id, or any identifier allowed by Git. subfolder (`str`, *optional*, defaults to `""`): The subfolder location of a model file within a larger model repository on the Hub or locally. mirror (`str`, *optional*): Mirror source to resolve accessibility issues if you're downloading a model in China. We do not guarantee the timeliness or safety of the source, and you should refer to the mirror site for more information. Example: To load a Textual Inversion embedding vector in 🤗 Diffusers format: ```py from diffusers import StableDiffusionPipeline import torch model_id = "stable-diffusion-v1-5/stable-diffusion-v1-5" pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16).to("cuda") pipe.load_textual_inversion("sd-concepts-library/cat-toy") prompt = "A backpack" image = pipe(prompt, num_inference_steps=50).images[0] image.save("cat-backpack.png") ``` To load a Textual Inversion embedding vector in Automatic1111 format, make sure to download the vector first (for example from [civitAI](https://civitai.com/models/3036?modelVersionId=9857)) and then load the vector locally: ```py from diffusers import StableDiffusionPipeline import torch model_id = "stable-diffusion-v1-5/stable-diffusion-v1-5" pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16).to("cuda") pipe.load_textual_inversion("./charturnerv2.pt", token="charturnerv2") prompt = "charturnerv2, multiple views of the same character in the same outfit, a character turnaround of a woman wearing a black jacket and red shirt, best quality, intricate details." image = pipe(prompt, num_inference_steps=50).images[0] image.save("character.png") ``` rBNr_rrKz&You have passed a state_dict contains z* embeddings, and list of tokens of length z% Make sure both have the same length.c3H># UHnTURS:gv M g7f)rN)ry).0embexpected_emb_dims r< ETextualInversionLoaderMixin.load_textual_inversion..sGJS399R=0Js"zLoaded embeddings are of incorrect shape. Expected each textual inversion embedding to be of shape {input_embeddings.shape[-1]}, but are {embeddings.shape[-1]} F_hf_hookhookszAccelerate hooks detected. Since you have called `load_textual_inversion()`, the previous hooks will be first removed. Then the textual inversion parameters will be loaded and the hooks will be applied again.)recursez'Loaded textual inversion embedding for rh)dtyper'r&)&getattrr)listr^rbr=r+r,rZrur~get_input_embeddingsweightryany hf_device_map componentsrmrModulehasattrrrrrrnrorr'rresize_token_embeddingsrj add_tokensconvert_tokens_to_idsdatatoenable_model_cpu_offloadenable_sequential_cpu_offload)rEr8r#rBr_r5r4rRr7r{is_model_cpu_offloadis_sequential_cpu_offloadrL componentr'rinput_embeddingsrttoken_idrs @r<r]2TextualInversionLoaderMixin.load_textual_inversionsnRA{D!A #Jwt^T'J ;TBB+ +. ' !+5$ 7 7%U !9 c"@AAF ##I=[d99Wb[ab  v;?s;/14+a.%,,77";q>2 v;#k"22$@[AQ@RR|~ABH~I}JJ?@ "AA&W`a"??T]^(<<>EEKKBO GJG G G_  %$)!    % $ 5 5 7 i33y*55/9')Z:XZd/e,&wy*'EGWXZ&y'9'97C Z *9+=+=+C+CA+FHX Y2  o58L8L3Ia!8 $$"" ,,S^c&k-IJ'<<>EE!$F 7 E9   ' 66u=H.7  ! !( + KKA%J K !8 %7  ) ) ) 8 &  . .f . ='rWrRcU=(d [USS5nU=(d [USS5n/nSnU(a[U[5(aU/nURR 5H=upgUR (d%UR U;aURU5 M9M;UnM? [U5S:Xa [S5eOf/nURR 5HFupgUR (d.URU5 URUR 5 MDUnMH [XA5HupURU URU M! Sn URH`nX:dM XU -:dMURUn XRXZ-'URU XZ-URU R 'U S- n Mb UR5 UR5 UR5R n UR5R"n U SUS-n/n[%US-U R&S5H.nUU;dM URU UR)S55 M0 [U5S:a+[*R,"USS9n[*R,"X/SS9n[.R0"UR&SU 5nUUR"lUR5U5 g)a Unload Textual Inversion embeddings from the text encoder of [`StableDiffusionPipeline`] Example: ```py from diffusers import AutoPipelineForText2Image import torch pipeline = AutoPipelineForText2Image.from_pretrained("stable-diffusion-v1-5/stable-diffusion-v1-5") # Example 1 pipeline.load_textual_inversion("sd-concepts-library/gta5-artwork") pipeline.load_textual_inversion("sd-concepts-library/moeb-style") # Remove all token embeddings pipeline.unload_textual_inversion() # Example 2 pipeline.load_textual_inversion("sd-concepts-library/moeb-style") pipeline.load_textual_inversion("sd-concepts-library/gta5-artwork") # Remove just one token pipeline.unload_textual_inversion("") # Example 3: unload from SDXL pipeline = AutoPipelineForText2Image.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0") embedding_path = hf_hub_download( repo_id="linoyts/web_y2k", filename="web_y2k_emb.safetensors", repo_type="model" ) # load embeddings to the text encoders state_dict = load_file(embedding_path) # load embeddings of text_encoder 1 (CLIP ViT-L/14) pipeline.load_textual_inversion( state_dict["clip_l"], tokens=["", ""], text_encoder=pipeline.text_encoder, tokenizer=pipeline.tokenizer, ) # load embeddings of text_encoder 2 (CLIP ViT-G/14) pipeline.load_textual_inversion( state_dict["clip_g"], tokens=["", ""], text_encoder=pipeline.text_encoder_2, tokenizer=pipeline.tokenizer_2, ) # Unload explicitly from both text encoders and tokenizers pipeline.unload_textual_inversion( tokens=["", ""], text_encoder=pipeline.text_encoder, tokenizer=pipeline.tokenizer ) pipeline.unload_textual_inversion( tokens=["", ""], text_encoder=pipeline.text_encoder_2, tokenizer=pipeline.tokenizer_2 ) ``` rBNr_rzNo tokens to remove foundrK)dim)rr)stradded_tokens_decoderrmspecialcontentr3r^rZrj_added_tokens_decoder_added_tokens_encoder _update_trie_update_total_vocab_sizer embedding_dimrrzry unsqueezer+catr Embeddingrset_input_embeddings)rErRrBr_ token_idslast_special_token_idadded_token_id added_tokenrtoken_to_removekey_idr#text_embedding_dimtemp_text_embedding_weightstext_embedding_weights to_appendrUtext_embeddings_filtereds r<unload_textual_inversion4TextualInversionLoaderMixin.unload_textual_inversions@A{D!A #Jwt^T'J  $ &#&& /8/M/M/S/S/U+"**"**f4!((85-;) 0V 9~" !<==#F/8/M/M/S/S/U+"**$$^4MM+"5"56,:) 0V*-Y)? %H//9//@*@ !66H/HW]?]4]!77ARW//0E0NO33H=AVA_ // >!  7  **,*>>@NN&2&G&G&I&P&P#!<=X?TWX?X!Y ,q02M2S2STU2VWA !  !A  )3I%*YY0F/RXY%Z "#%<<0F0L0LQ0OQc#d /E '',))*BCrW)NNN)r\ __module__ __qualname____firstlineno____doc__rrrrHrDrb staticmethodrur~rrr+r,rr]r__static_attributes__rrWr<r?r?nsg5d3i+@Mb6C',S$s)T#u||BS=TVZ[_`cejeqeq`q[rVs-s't>c49n-.>12 > 01 >>J375948 zDsDI~./zD12zD01 zDzDrWr?)typingrrrrr/r+huggingface_hub.utilsrrmodels.modeling_utilsr utilsr r r r transformersrraccelerate.hooksrrr get_loggerr\rnr2r.r=r?rrWr<rs/. 63``AVV   H %*7EEPWDWDrW