+hZSSKrSSKJr SSKJr SSKJr SSKJrJ r J r SSK r SSK r SSK Js Jr SSKJr SSKJrJrJrJrJrJr SS KJr SS KJrJr SS KJ r J!r!J"r"J#r#J$r$J%r%J&r&J'r'J(r(J)r) SS K*J+r+ S SK,J-r- S SK.J/r/J0r0J1r1J2r2 S SKJ3r3 \)Rh"\55r6Sr7Sr8"SS5r9g)N) defaultdict) nullcontext)Path)CallableDictUnion)validate_hf_hub_args)ImageProjectionIPAdapterFaceIDImageProjection"IPAdapterFaceIDPlusImageProjectionIPAdapterFullImageProjectionIPAdapterPlusImageProjectionMultiIPAdapterImageProjection)load_model_dict_into_meta)_LOW_CPU_MEM_USAGE_DEFAULTload_state_dict) USE_PEFT_BACKEND_get_model_fileconvert_unet_state_dict_to_peft deprecateget_adapter_nameget_peft_kwargsis_accelerate_availableis_peft_versionis_torch_versionlogging)empty_device_cache#_func_optionally_disable_offloading)LORA_WEIGHT_NAMELORA_WEIGHT_NAME_SAFETEXT_ENCODER_NAME UNET_NAME)AttnProcsLayersz$pytorch_custom_diffusion_weights.binz,pytorch_custom_diffusion_weights.safetensorsc \rSrSrSr\r\r\ S\ \ \ \ \ R444Sj5rSrSr\S5rSS \ \ \R*4S \S \ S \S \4 SjjrSr\4Sjr\4Sjr\4SjrSrSrg)UNet2DConditionLoadersMixin<z2 Load LoRA layers into a [`UNet2DCondtionModel`]. %pretrained_model_name_or_path_or_dictc SSKJn URSS5nURSS5nURSS5nURSS5nURS S5nURS S5n URS S5n URS S5n URS S5n URSS5n URSS5nURSS5nURS[5nSnU(a[ SS5(a [ S5eU cSn SnSSS.nSn[ U[5(dU (aU bU bTU RS5(a>[UU =(d [UUUUUU U US9 n[RRUSS9nUc*[UU =(d [UUUUUU U US9 n[!U5nOUn[#SWR%555n['SUR%555nSnSnSnU(aSn[)S S!U5 U(aUR+US"9nO8U(a#UR-UUR.UU UUS#9unnnO[ US$35eU(aKUbHUR1US%9unnnUR3W5 UR5UR6UR8S&9 U(aUR;5 gU(aUR=5 gU(aUUR>RA5H6n[ U[RBRD5(dM.U"U5 M8 gg![anU(dUeSnAGNSnAff=f)'a Load pretrained attention processor layers into [`UNet2DConditionModel`]. Attention processor layers have to be defined in [`attention_processor.py`](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py) and be a `torch.nn.Module` class. Currently supported: LoRA, Custom Diffusion. For LoRA, one must install `peft`: `pip install -U peft`. Parameters: pretrained_model_name_or_path_or_dict (`str` or `os.PathLike` or `dict`): Can be either: - A string, the model id (for example `google/ddpm-celebahq-256`) of a pretrained model hosted on the Hub. - A path to a directory (for example `./my_model_directory`) containing the model weights saved with [`ModelMixin.save_pretrained`]. - A [torch state dict](https://pytorch.org/tutorials/beginner/saving_loading_models.html#what-is-a-state-dict). 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. 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. network_alphas (`Dict[str, float]`): The value of the network alpha used for stable learning and preventing underflow. This value has the same meaning as the `--network_alpha` option in the kohya-ss trainer script. Refer to [this link](https://github.com/darkstorm2150/sd-scripts/blob/main/docs/train_network_README-en.md#execute-learning). adapter_name (`str`, *optional*, defaults to None): Adapter name to be used for referencing the loaded adapter model. If not specified, it will use `default_{i}` where i is the total number of adapters being loaded. weight_name (`str`, *optional*, defaults to None): Name of the serialized state dict file. low_cpu_mem_usage (`bool`, *optional*): Speed up model loading by only loading the pretrained LoRA weights and not initializing the random weights. Example: ```py from diffusers import AutoPipelineForText2Image import torch pipeline = AutoPipelineForText2Image.from_pretrained( "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16 ).to("cuda") pipeline.unet.load_attn_procs( "jbilcke-hf/sdxl-cinematic-1", weight_name="pytorch_lora_weights.safetensors", adapter_name="cinematic" ) ``` r )*_maybe_remove_and_reapply_group_offloading cache_dirNforce_downloadFproxieslocal_files_onlytokenrevision subfolder weight_nameuse_safetensors adapter_name _pipelinenetwork_alphaslow_cpu_mem_usage<=z0.13.0zq`low_cpu_mem_usage=True` is not compatible with this `peft` version. Please update it with `pip install -U peft`.Tattn_procs_weightspytorch) file_type frameworkz .safetensors) weights_namer-r.r/r0r1r2r3 user_agentcpu)devicec3,# UH nSU;v M g7f)custom_diffusionN.0ks P/home/james-whalen/.local/lib/python3.13/site-packages/diffusers/loaders/unet.py >UNet2DConditionLoadersMixin.load_attn_procs..s!UCTa"4"9CTc3\# UH"nSU;=(d URS5v M$ g7f)loraz.alphaN)endswithrFs rIrJrKs'WEVv{:ajj&::EVs*,zUsing the `load_attn_procs()` method has been deprecated and will be removed in a future version. Please use `load_lora_adapter()`.load_attn_procs0.40.0) state_dict)rRunet_identifier_keyr8r6r7r9zQ does not seem to be in the correct format expected by Custom Diffusion training.r7dtyperB)#hooks.group_offloadingr,poprr ValueError isinstancedictrOrr# safetensorstorch load_fileIOErrorr"ranykeysallr_process_custom_diffusion _process_lora unet_name_optionally_disable_offloadingset_attn_processortorVrBenable_model_cpu_offloadenable_sequential_cpu_offload componentsvaluesnnModule)selfr*kwargsr,r-r.r/r0r1r2r3r4r5r6r7r8r9 allow_pickler@ model_filerReis_custom_diffusionis_lorais_model_cpu_offloadis_sequential_cpu_offloadis_group_offloaddeprecation_messageattn_processors components rIrP+UNet2DConditionLoadersMixin.load_attn_procsDsH XJJ{D1 $4e<**Y-!::&8$? 7D)::j$/JJ{D1 jj5  **%6=zz.$7 JJ{D1 $4d;"JJ':J!!U:??CT!UUWZ__EVWW$$)!  #h  '3F G "<< <SO PTPbPb%$(NN-)#"3 QcQ M ";=M,op  9#8PTPsPs#QtQ M ";=M  # #O 4 GG$**T[[G 9  . . 0 &  3 3 5 &1188: i99>yI;G' s!=M M6" M11M6cSSKJn 0n[[5nUR 5HupV[ U5S:Xa0XE'MSU;aFSR URS5SS5SR URS5SS5pOESR URS5SS5SR URS5SS5pXdUU'M UR 5HmupY[ U 5S:Xa U"SSSSS 9X5'M!U S RS n U S RSn S U ;aS OSn U"S U U U S 9X5'X5RU 5 Mo U$)Nr )CustomDiffusionAttnProcessorrto_out.F)train_kv train_q_out hidden_sizecross_attention_dimzto_k_custom_diffusion.weightrzto_q_custom_diffusion.weightT) models.attention_processorr~rr[itemslenjoinsplitshaper) rorRr~rzcustom_diffusion_grouped_dictkeyvalueattn_processor_keysub_key value_dictrrrs rIrc5UNet2DConditionLoadersMixin._process_custom_diffusionsM(3D(9%$**,JC5zQ57-2s?25((399S>#2;N2OQTQYQYZ]ZcZcdgZhikilZmQn25((399S>#2;N2OQTQYQYZ]ZcZcdgZhikilZmQnMR.@A'J- =BBDOC:!#'C"4]a($'11O&P&V&VWX&Y#()GHNNqQ &D &RdX] 'C! + +(; ($  $44Z@ E"c@[(d [S5eSSKJnJnJn [ UR55n U V s/sHoRU5(dMU PM n n UR5V V s0sH"upX;dM U RUS3S5U _M$ nn n UbvUR5V s/sHoRU5(dMU PM nn UR5V V s0sH"upX;dM U RUS3S5U _M$ nn n SnSnSn[U5S:aUOUn[U5S:Ga7U[US05;a[SUS 35e[U5nUb [U5n0nUR5H unnS U;dMURS UU'M" [UX1S S 9nSU;aIUS(a[!SS5(a [S5eO"[!SS5(aUR#S5 SU;aIUS(a[!SS5(a [S5eO"[!SS5(aUR#S5 U"S!0UD6nUc [%U5nUR'U5unnn0n[!SS5(aUUS'U"UU4SU0UD6 U "XU40UD6nSnUb[USS5nU(a=UV s/sHn SU ;dM XK;dMU PM nn U(aSSR)U5S3n[USS5nU(a@UV s/sHn SU ;dM XK;dMU PM nn U(aUS SR)U5S3- nU(a[*R-U5 UUU4$s sn fs sn n fs sn fs sn n fs sn fs sn f)"Nz)PEFT backend is required for this method.r) LoraConfiginject_adapter_in_modelset_peft_model_state_dictrF peft_configz Adapter name z? already in use in the Unet - please select a new adapter name.lora_BrT)is_unetuse_dora=z0.13.1r9r6unexpected_keyslora_zSLoading adapter weights from state_dict led to unexpected keys found in the model: z, z. missing_keyszJLoading adapter weights from state_dict led to missing keys in the model: rE)rrYpeftrrrlistra startswithrreplacergetattrrrrrrXrrfrloggerwarning) rorRrSr8r6r7r9rrrrarH unet_keysvunet_state_dict alpha_keysrvrwrxstate_dict_to_be_usedrankrvallora_config_kwargs lora_config peft_kwargsincompatible_keyswarn_msgrlora_unexpected_keysrlora_missing_keyss rIrd)UNet2DConditionLoadersMixin._process_lora#s HI IWWJOO%& $J1 5H(IQ JDNDTDTDV DVDAZ[Zh 7AII,-Q/ 4a 7DV    %%3%8%8%:`%:llK^>_!%:J`HVH\H\H^H^bcbq; 013R8!;H^  %$)! 363G!3KQ[ $ % )wt]B?? #L>1pq99NOJ)"A!PD&,,.Ss? # ! DI/"1~[_!` //%j1&sG44(D5 'sG44*..z:00%k2&tX66(B7 'tX66*..{;$:'9:K#/5 QUPsPsQ M ";=MKtX..3D /0 #K `L `T_ ` 9$L h\g h H ,")*;=NPT"U"7F+m!'UV,A[g[lA(+m+ $ *> ?@D! ''8.$O 4@(gLqGqLUaUfL%(g(  $ *; <=Q@ x(#%>@PPPEK aN,n)hsSNN6 NN7N N / N>N N N'N# N1N8Nc[US9$)NrTr )clsr7s rIrf:UNet2DConditionLoadersMixin._optionally_disable_offloadings3YGGrNsave_directoryis_main_processr4 save_functionsafe_serializationc r^^^SSKJmJmJm [R R U5(a[RSUS35 g[UUU4SjURR555nU(aUR5nUcU(aUR5V V s0sH(up[U [R5(aM&X_M* n n n [!U 5S:a&[R#SU R%535 UR5V V s0sH(up[U [R5(dM&X_M* nn n O3S n ['S S U 5 [((d [+S 5eSS KJn U "U5nUcU(aSnO[R0n[R2"USS9 Uc.U(aU(a[4O[6nOU(a[8O[:n[=X5R?5nU"X5 [RASU35 gs sn n fs sn n f)a Save attention processor layers to a directory so that it can be reloaded with the [`~loaders.UNet2DConditionLoadersMixin.load_attn_procs`] method. Arguments: save_directory (`str` or `os.PathLike`): Directory to save an attention processor to (will be created if it doesn't exist). is_main_process (`bool`, *optional*, defaults to `True`): Whether the process calling this is the main process or not. Useful during distributed training and you need to call this function on all processes. In this case, set `is_main_process=True` only on the main process to avoid race conditions. save_function (`Callable`): The function to use to save the state dictionary. Useful during distributed training when you need to replace `torch.save` with another method. Can be configured with the environment variable `DIFFUSERS_SAVE_MODE`. safe_serialization (`bool`, *optional*, defaults to `True`): Whether to save the model using `safetensors` or with `pickle`. Example: ```py import torch from diffusers import DiffusionPipeline pipeline = DiffusionPipeline.from_pretrained( "CompVis/stable-diffusion-v1-4", torch_dtype=torch.float16, ).to("cuda") pipeline.unet.load_attn_procs("path-to-save-model", weight_name="pytorch_custom_diffusion_weights.bin") pipeline.unet.save_attn_procs("path-to-save-model", weight_name="pytorch_custom_diffusion_weights.bin") ``` r r~CustomDiffusionAttnProcessor2_0$CustomDiffusionXFormersAttnProcessorzProvided path (z#) should be a directory, not a fileNc3F># UHup[UTTT45v M g7f)N)rZ)rG_xr~rrs rIrJ>UNet2DConditionLoadersMixin.save_attn_procs..s7" 7 -/NPtu  7s!rzfSafetensors does not support saving dicts with non-tensor values. The following keys will be ignored: zUsing the `save_attn_procs()` method has been deprecated and will be removed in a future version. Please use `save_lora_adapter()`.save_attn_procsrQzOPEFT backend is required for saving LoRAs using the `save_attn_procs()` method.)get_peft_model_state_dictcB[RRXSS0S9$)Nformatpt)metadata)r\r] save_file)weightsfilenames rIrBUNet2DConditionLoadersMixin.save_attn_procs..save_functions%&,,66wT\^bSc6ddrT)exist_okzModel weights saved in )!rr~rrospathisfilererrorr`rzr _get_custom_diffusion_state_dictrZr]TensorrrrarrrY peft.utilsrsavemakedirs!CUSTOM_DIFFUSION_WEIGHT_NAME_SAFEr#CUSTOM_DIFFUSION_WEIGHT_NAMEr"ras_posixinfo)rorrr4rrrprtrRrHrempty_state_dictryr save_pathr~rrs @@@rIr+UNet2DConditionLoadersMixin.save_attn_procssR  77>>. ) ) LL?>*::]^ _ !" ..446 "   >>@J$);5?5E5E5G#k5GTQzZ[]b]i]iOjDAD5G #k'(1,NN??O?T?T?V>WY0:/?/?/Aa/AtqZPQSXS_S_E`dad/A a#h  '3F G## !rss <248J  !e!&  NT2  !CV?\q >Q:Wg 5>>@ j, -i[9:M$l bs/%H-H-)%H3H3c nSSKJnJnJn [ UR R 5VVs0sHupE[UUUU45(dMXE_M snn5nUR5nUR R 5H(up[U R55S:XdM$0Xx'M* U$s snnf)Nr rr) rr~rrr&rzrrZrRr) ror~rryr model_to_saverRnameattns rIrS?5n U RS@SA5n U RSBSC5n U RSDSE5n U RSFSG5n U RSHSI5n SJU ;aXU RSKSL5'GMSMU ;aXU RSNSO5'GMSPU ;aAU RSQSSR9nUSXLRSPSS5'USXLRSPST5'GMSUU ;aXU RSUSV5'GM0SU :XaXSW'GM=SXU :XaXSY'GMJSZU :XaXS['GMWS\U :XaXS]'GMdXU 'GMk GOS^U;aUSRSnUSRSnUU-nS^nUURSn USZRSU -nU"5 [U UUUS_9nSSS5 UR5H-upU RSS5n U RSS5n XU 'M/ GO'US`RSnUSaRSnUSbRSnUS`RSQn[!ScU55nU(aUSdRSS-OUSeRSS-nU"5 [#UUUUUSf9nSSS5 UR5GHjupU RS Sg5n U RShSi5n U RSjSk5n U RSlSm5n U RSnSo5n U RSpSq5n U RSrSs5n U RStSu5n U RSvSw5n SPU ;ahU R%Sx5nSyUSQ'SxR'U5n U RSQSSR9nUSXLRSPSS5'USXLRSPST5'GMSzU ;a.U R%Sx5nSyUSQ'SxR'U5n XU 'GMJSUU ;a>U R%Sx5nSyUSQ'SxR'U5n XU RSUSV5'GMU RS{S|5n U RS}S~5n U RSS5n U RSS5n U RSS5n U RSS5n U RSS5n U RSS5n U RSS5n U RSS5n U RSS5n U RSS5n XU 'GMm U(dUR)USS9 U$SUR*0n[-XTUUR.S9 [15 U$!,(df  GN=f!,(df  GN=f!,(df  GN=f!,(df  GN=f!,(df  GN=f)Nrinit_empty_weightsF,Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with: ``` pip install accelerate ``` .Tr1.9.0~Low memory initialization requires torch >= 1.9.0. Please either update your PyTorch version or set `low_cpu_mem_usage=False`. proj.weight)rimage_embed_dimnum_image_text_embedsproj image_embeds proj.3.weightz proj.0.weight)rrzproj.0z ff.net.0.projzproj.2zff.net.2zproj.3norm"perceiver_resampler.proj_in.weightrz#perceiver_resampler.proj_out.weightz*perceiver_resampler.layers.0.0.to_q.weight@) embed_dims output_dims hidden_dimsheadsid_embeddings_dimzperceiver_resampler.rz0.tozattn.toz0.1.0.z0.ff.0.z 0.1.1.weightz0.ff.1.net.0.proj.weightz 0.1.3.weightz0.ff.1.net.2.weightz1.1.0.z1.ff.0.z 1.1.1.weightz1.ff.1.net.0.proj.weightz 1.1.3.weightz1.ff.1.net.2.weightz2.1.0.z2.ff.0.z 2.1.1.weightz2.ff.1.net.0.proj.weightz 2.1.3.weightz2.ff.1.net.2.weightz3.1.0.z3.ff.0.z 3.1.1.weightz3.ff.1.net.0.proj.weightz 3.1.3.weightz3.ff.1.net.2.weightz layers.0.0z layers.0.ln0z layers.0.1z layers.0.ln1z layers.1.0z layers.1.ln0z layers.1.1z layers.1.ln1z layers.2.0z layers.2.ln0z layers.2.1z layers.2.ln1z layers.3.0z layers.3.ln0z layers.3.1z layers.3.ln1norm1z0.norm10norm2z0.norm21to_kvr )dimto_kto_vrzto_out.0zproj.net.0.proj.weightz proj.0.biaszproj.net.0.proj.biasz proj.2.weightzproj.net.2.weightz proj.2.biaszproj.net.2.bias norm.weight)rrmult num_tokenslatentszproj_in.weightzproj_out.weightc3,# UH nSU;v M g7f)rNrErFs rIrJZUNet2DConditionLoadersMixin._convert_ip_adapter_image_proj_to_diffusers..s"C 16Q; rLzlayers.0.attn.to_q.weightzlayers.0.0.to_q.weight)rrrr num_queriesz2.toz 0.0.norm1z0.ln0z 0.0.norm2z0.ln1z 1.0.norm1z1.ln0z 1.0.norm2z1.ln1z 2.0.norm1z2.ln0z 2.0.norm2z2.ln1z 3.0.norm1z3.ln0z 3.0.norm2z3.ln1rrto_qz0.1.0z0.ff.0z0.1.1z0.ff.1.net.0.projz0.1.3z 0.ff.1.net.2z1.1.0z1.ff.0z1.1.1z1.ff.1.net.0.projz1.1.3z 1.ff.1.net.2z2.1.0z2.ff.0z2.1.1z2.ff.1.net.0.projz2.1.3z 2.ff.1.net.2z3.1.0z3.ff.0z3.1.1z3.ff.1.net.0.projz3.1.3z 3.ff.1.net.2)strict device_maprV)r acceleraterrrrNotImplementedErrorrrr rrrr chunkr r`rrrrrBrrVr)rorRr9rupdated_state_dictimage_projection init_contextrclip_embeddings_dimrrrdiffusers_namerrrrrv_chunkid_embeddings_dim_inid_embeddings_dim_out multiplier norm_layerr attn_key_presentpartsrs rI+_convert_ip_adapter_image_proj_to_diffusersGUNet2DConditionLoadersMixin._convert_ip_adapter_image_proj_to_diffuserss &((9%*!2  $-=dG-L-L%.   ->)K J &$% !",]";"A"A""E ",]";"A"A!"D"I #2(;$7*?$  )..0 !$V^!D5:>21 *",_"="C"CA"F ",_"="C"CA"F #?(;$  )..0 !$X!G!/!7!7*!M!/!7!7&!I5:>2 1 2Z ? *? ; A A! D #$HIOOPQRJ$%IJPPQRSK$%JKQQRSTKKLRRSTUY[[E#E) + +&7 $  )..0 !$-CR!H!/!7!7 !J!/!7!7)!L!/!7!7Hb!c!/!7!7H]!^!/!7!7)!L!/!7!7Hb!c!/!7!7H]!^!/!7!7)!L!/!7!7Hb!c!/!7!7H]!^!/!7!7)!L!/!7!7Hb!c!/!7!7H]!^!/!7!7 n!U!/!7!7 n!U!/!7!7 n!U!/!7!7 n!U!/!7!7 n!U!/!7!7 n!U!/!7!7 n!U!/!7!7 n!Un,QV~'='=i'MN.QV~'='=i'MN.#kk!k3GRYZ[R\&'='=gv'NORYZ[R\&'='=gv'NO/W\~'='=h 'ST$6CH'?@"n4AF'=>$6>C':;"n4~6W1Zj (#-o#>#D#DQ#G $.$?$E$Ea$H !.2FFJ&J",Z"8">">q"A #O4::1=ATTJ#A(;$8#) $  )..0 !$X!G!/!7!7*!M5:>21%/y$9$?$?$B !#$45;;A>J$%67==a@K$Y/55a8K""C "CC $67==a@BF 89??BbH  #?) + + 5 $  )..0 !$VV!<!/!7!7 W!M!/!7!7 W!M!/!7!7 W!M!/!7!7 W!M!/!7!7 W!M!/!7!7 W!M!/!7!7 W!M!/!7!7 W!Mn,*005E%E!H%(XXe_N#kk!k3GRYZ[R\&'='=gv'NORYZ[R\&'='=gv'NO~-*005E%E!H%(XXe_N9>~6/*005E%E!H%(XXe_NW\~'='=h 'ST%3%;%;GX%NN%3%;%;GEX%YN%3%;%;G^%TN%3%;%;GX%NN%3%;%;GEX%YN%3%;%;G^%TN%3%;%;GX%NN%3%;%;GEX%YN%3%;%;G^%TN%3%;%;GX%NN%3%;%;GEX%YN%3%;%;G^%TN9>~6[1^!  , ,-? , M   dkk*J %&6Waimisis t  w   & ~ 4 s< b" b4'c cc*" b14 c c c'* c9c  SSKJnJnJn U(a-[ 5(aSSKJn OSn[RS5 USLa[SS 5(d [S 5e0nS nU(aWO[n URR5GHn U RS 5(aSOURR n U R#S 5(aURR$Sn OU R#S5(aC['U [)S55n [+[-URR$55U n OFU R#S5(a0['U [)S55n URR$U n U bSU ;a%URU R.nU"5Xz'GM/S[1URU R.5;aUnO[3[4S5(aUOUn/nUH`nSUS;aUS/- nMSUS;aUS/- nM%SUS;aUS/- nM6SUS;aUS/- nMGUUSSR6S /- nMb U "5 U"W U SUS9Xz'SSS5 0n[9U5HFunnUR;S US!3US"US#305 UR;S$US!3US"US%305 MH U(dXzR=U5 Oj[?[AURC555RDn[?[AURC555RFnS&U0n[IXzUUUS'9 US- nGM [K5 U$!,(df  GN=f)(Nr )IPAdapterAttnProcessorIPAdapterAttnProcessor2_0IPAdapterXFormersAttnProcessorrrFrTrrrrzattn1.processor mid_blockr up_blocksz up_blocks. down_blocksz down_blocks.motion_modulesXFormersscaled_dot_product_attentionr image_projrrirrr g?)rrscaler zto_k_ip.z.weight ip_adapterz.to_k_ip.weightzto_v_ip.z.to_v_ip.weightrr)&rr%r&r'rrrrrrrrrzrarOconfigrrblock_out_channelsintrrreversed __class__strhasattrFr enumerateupdaternextiterrlrBrVrr)ro state_dictsr9r%r&r'r attn_procskey_idrrrrblock_idattn_processor_classrrRrirBrVrs rI%_convert_ip_adapter_attn_to_diffusersAUNet2DConditionLoadersMixin._convert_ip_adapter_attn_to_diffuserss  &((9%*!2  $-=dG-L-L%.   ->)K ((--/D*.--8I*J*J$PTP[P[PoPo {++"kk<$.F'+';';D'A'K'K$#7#9  T%9%9$%?%I%I!JJ+I(#1&DEE23) )+%"-J$ <(@@-!4-(J|,DD-#6-=LAYY-!4-&*\*BB-!4-.*\2J92U2[2[\]2^1__-#.""^';$/,?!#8 (J$$ %.{%;MAz%%!G'VZ`YaapWq>r&st%%!G'VZ`YaapWq>r&st&<)$44Z@!$z'8'8':";<CCF j&7&7&9!:;AAE"$fJ-j.> Waino! }0@ 3$^s - M33 N c[U[5(dU/nURb<URRS:Xa"[ US5(dURUlSUlURXS9nURU5 /nUH'nURUSUS9nURU5 M) [U5UlSURlURURURS9 g)N text_projtext_encoder_hid_proj)r9r. ip_image_projrU)rZrencoder_hid_projr1encoder_hid_dim_typer7rGrCrgr"appendrrhrVrB)ror=r9r>image_projection_layersrRimage_projection_layers rI_load_ip_adapter_weights4UNet2DConditionLoadersMixin._load_ip_adapter_weights\s+t,,&-K  ! ! - 00K?D"9::)-)>)>D &!%?? ?q   +#%%J%)%U%U<(>U V+: ( djj5rc0n[URR55GHCup4[U5GH-upVUS3US;dMXR;a0X%'X%RSUS3USUS305 X%RSUS3USUS305 X%RSUS3USUS305 X%RSUS3USUS305 X%RSUS3USUS305 X%RSUS3USUS305 X%RSUS 3USUS 305 X%RSUS 3USUS 305 GM0 GMF U$) Nz.to_k_lora.down.weightr0zunet.z.to_q_lora.down.weightz.to_v_lora.down.weightz.to_out_lora.down.weightz.to_k_lora.up.weightz.to_q_lora.up.weightz.to_v_lora.up.weightz.to_out_lora.up.weight)r9rzrar:)ror= lora_dictsr?rrBrRs rI_load_ip_adapter_loras2UNet2DConditionLoadersMixin._load_ip_adapter_loras|s7 %d&:&:&?&?&ABLF!*;!7 X34 <8PP*(* M((#D6)?@*\BZ#)(*@ ACM((#D6)?@*\BZ#)(*@ ACM((#D6)?@*\BZ#)(*@ ACM((#D6)ABJ|D\#)(*B CEM(( &:;Z =UY_X``tVu=vwM(( &:;Z =UY_X``tVu=vwM(( &:;Z =UY_X``tVu=vwM((#D6)?@*\BZ#)(*@ ACS"8Cbr)rIrG)TNNT) __name__ __module__ __qualname____firstlineno____doc__r$text_encoder_namer%rer rr6rr]rrPrcrd classmethodrfrPathLikeboolrrrrr"rCrNrR__static_attributes__rErrIr(r(<s*IyJU3PTUXZ_ZfZfUfPgKgEhyJyJx DsQjHH !%"&#' d;c2;;./d;d; d;  d; ! d;L8YsY vTn`DGa6@3rr():r collectionsr contextlibrpathlibrtypingrrrr\r]torch.nn.functionalrm functionalr8huggingface_hub.utilsr models.embeddingsr r r rrrmodels.model_loading_utilsrmodels.modeling_utilsrrutilsrrrrrrrrrrutils.torch_utilsr lora_baser! lora_pipeliner"r#r$r%r& get_loggerrTrrrr(rErrIrms #"(( 6CO   3:``"   H % F$R!s s r