+h BSSKJr SSKJrJrJr SSKrSSKJr SSKJ r J r SSK J r J r SS KJr SS KJrJrJr SS KJr SS KJr S SKJrJr \ R6"\5r\"SS\ 55r"SS\\ 5r"SS\5r g)) dataclass)DictOptionalUnionN)nn) ConfigMixinregister_to_config) BaseOutputlogging)AttentionProcessor)-HunyuanCombinedTimestepTextSizeStyleEmbedding PatchEmbedPixArtAlphaTextProjection) ModelMixin)HunyuanDiTBlock)Tuple zero_modulec:\rSrSr%\\R \S'Srg)HunyuanControlNetOutput$controlnet_block_samplesN) __name__ __module__ __qualname____firstlineno__rtorchTensor__annotations____static_attributes__ri/home/james-whalen/.local/lib/python3.13/site-packages/diffusers/models/controlnets/controlnet_hunyuan.pyrr$s#ELL11r$rc4^\rSrSr\SS\S\S\S\\S\\S\S\S \S \S \S \S \S\S\ 4U4Sjjj5r \ S\ \\ 44Sj5rS\\ \ \\ 444Sjr\SSj5rSS\R(S\4SjjrSrU=r$)HunyuanDiT2DControlNetModel)conditioning_channelsnum_attention_headsattention_head_dim in_channels patch_size activation_fntransformer_num_layers mlp_ratiocross_attention_dimcross_attention_dim_t5pooled_projection_dimtext_len text_len_t5"use_style_cond_and_image_meta_sizecr>[TU]5 X lX#-Ul[ U U S-U SS9Ul[ R"[R"X-U [RS95Ul [UUUUUSS9Ul [UU UU US9Ul[ R "/5Ul[ R "[%U S-S- 5Vs/sHFn['URUR(R*U[-URU -5U S S S 9PMH sn5Ul[1[ R2"X55Ul[%[7UR.55H?n[ R2"X5n[1U5nUR"R9U5 MA gs snf) N silu_fp32) in_features hidden_size out_featuresact_fn)dtype)heightwidthr, embed_dimr-pos_embed_type)r3seq_lenr1r6r rTF)dimr*r. ff_inner_dimr1qk_normskip)super__init__ num_heads inner_dimr text_embedderr Parameterr randnfloat32text_embedding_paddingr pos_embedrtime_extra_emb ModuleListcontrolnet_blocksrangerconfigr*intblocksrLinear input_blocklenappend)selfr)r*r+r,r-r. sample_sizer;r/r0r1r2r3r4r5r6layer_controlnet_block __class__s r%rI$HunyuanDiT2DControlNetModel.__init__*s( ,,A6..2,  ')ll KK.0C5== Y' #$#!!  L "7 6/Q  "$r!2mm##9Q#>#BC DE (, (G(G"/!$T^^i%?!@(; D    'ryy'JKs4;;'(A!yyB *+;<   " " ) )*: ;) sA F4returnc^0nS[S[RRS[[[ 44U4SjjmUR 5Hup#T"X#U5 M U$)z Returns: `dict` of attention processors: A dictionary containing all attention processors used in the model with indexed by its weight name. namemodule processorsc>[US5(aURSS9X S3'UR5Hup4T"USU3XB5 M U$)N get_processorT)return_deprecated_lora .processor.)hasattrrjnamed_children)rfrgrhsub_namechildfn_recursive_add_processorss r%rrPHunyuanDiT2DControlNetModel.attn_processors..fn_recursive_add_processorss`v//282F2F^b2F2c V:./#)#8#8#:+tfAhZ,@%T$; r$)strr rModulerrro)r]rhrfrgrrs @r%attn_processors+HunyuanDiT2DControlNetModel.attn_processorsvsb  c 588?? X\]`bt]tXu !//1LD 'j A2r$ processorc d^[URR55n[U[5(a-[U5U:wa[ S[U5SUSUS35eS[ S[RR4U4SjjmUR5Hup4T"X4U5 M g) a Sets the attention processor to use to compute attention. Parameters: processor (`dict` of `AttentionProcessor` or only `AttentionProcessor`): The instantiated processor class or a dictionary of processor classes that will be set as the processor for **all** `Attention` layers. If `processor` is a dict, the key needs to define the path to the corresponding cross attention processor. This is strongly recommended when setting trainable attention processors. z>A dict of processors was passed, but the number of processors z0 does not match the number of attention layers: z. Please make sure to pass z processor classes.rfrgc >[US5(aJ[U[5(dURU5 O#URUR US355 UR 5Hup4T"USU3XB5 M g)N set_processorrlrm)rn isinstancedictr{popro)rfrgrxrprqfn_recursive_attn_processors r%rSHunyuanDiT2DControlNetModel.set_attn_processor..fn_recursive_attn_processorssv//!)T22((3(($z7J)KL#)#8#8#:+tfAhZ,@%S$;r$N) r[rvkeysr|r} ValueErrorrtr rruro)r]rxcountrfrgrs @r%set_attn_processor.HunyuanDiT2DControlNetModel.set_attn_processorsD((--/0 i & &3y>U+BPQTU^Q_P`a005w6QRWQXXkm   Tc T588?? T!//1LD 'i @2r$cURnURnURnURnURn UR n UR n URn URn URnURnURnURnUnU=(d URnU"UUUUUU U U U U UUUUS9nU(a9URUR5SS9n[ R#SUS35 U$)N)r)r/r.r+r1r2r;r,r0r*r-r^r4r5F)strictz0controlnet load from Hunyuan-DiT. missing_keys: r)rVr.r+r1r2r;r,r0r*r-r^r4r5r/load_state_dict state_dictloggerwarning)cls transformerr)r/load_weights_from_transformerrVr.r+r1r2r;r,r0r*r-r^r4r5 controlnetkeys r%from_transformer,HunyuanDiT2DControlNetModel.from_transformers"##,, #66$88!'!>!>(( (( $$ $88&& (( ??(( 5!7!X6;X;X"7#9'1 3#9## 3!## ),,[-C-C-Ee,TC NNMcRSfXV Wr$controlnet_condconditioning_scalec TURSSupURU5nXRURU55-nURX'XURS9nURunnnUR UR SURS55nUR UUS5n[R"XW/SS9n[R"Xh/SS9nURS5R5n[R"XeUR5nSn[UR5HunnU"UUUU S 9nUU4-nM Sn[UUR 5HunnU"U5nUU4-nM UVs/sHnUU-PM nnU (dU4$[#US 9$s snf)  The [`HunyuanDiT2DControlNetModel`] forward method. Args: hidden_states (`torch.Tensor` of shape `(batch size, dim, height, width)`): The input tensor. timestep ( `torch.LongTensor`, *optional*): Used to indicate denoising step. controlnet_cond ( `torch.Tensor` ): The conditioning input to ControlNet. conditioning_scale ( `float` ): Indicate the conditioning scale. encoder_hidden_states ( `torch.Tensor` of shape `(batch size, sequence len, embed dims)`, *optional*): Conditional embeddings for cross attention layer. This is the output of `BertModel`. text_embedding_mask: torch.Tensor An attention mask of shape `(batch, key_tokens)` is applied to `encoder_hidden_states`. This is the output of `BertModel`. encoder_hidden_states_t5 ( `torch.Tensor` of shape `(batch size, sequence len, embed dims)`, *optional*): Conditional embeddings for cross attention layer. This is the output of T5 Text Encoder. text_embedding_mask_t5: torch.Tensor An attention mask of shape `(batch, key_tokens)` is applied to `encoder_hidden_states`. This is the output of T5 Text Encoder. image_meta_size (torch.Tensor): Conditional embedding indicate the image sizes style: torch.Tensor: Conditional embedding indicate the style image_rotary_emb (`torch.Tensor`): The image rotary embeddings to apply on query and key tensors during attention calculation. return_dict: bool Whether to return a dictionary. N) hidden_dtyper)rDr r)tembencoder_hidden_statesimage_rotary_emb)r)shaperQrZrRr>rLviewr cat unsqueezeboolwhererP enumeraterXziprTr)r] hidden_statestimesteprrrtext_embedding_maskencoder_hidden_states_t5text_embedding_mask_t5image_meta_sizestyler return_dictr?r@r batch_sizesequence_lengthr`block_res_samplesr_blockcontrolnet_block_res_samplesblock_res_samplerasamples r%forward#HunyuanDiT2DControlNetModel.forwards^&++BC0 }5 &(8(89X(YY "" U]UcUc#  *B)G)G& OQ#'#5#5 $ ) )".F.L.LR.P Q$  $<#@#@_^`#a % +@*[ab c#ii)<(U[]^1;;A>CCE % ,?X\XsXs t%dkk2LE5!&;!1 M!2]4D D 3(*$256GI_I_2` . ./0@A +GK[J]+] (3a So'oRn1C(CRn$'o02 2&@\]] (psF%) rXrTrKrZrJrQrLrPrR)rXNNzgelu-approximate i(g@irMT)rNT g?NNNNNNNT)rrrrr rWrrtfloatrrIpropertyrrrvrr classmethodrr r!rr# __classcell__rbs@r%r'r')s&'#%"$%)$(/&(#'&*%)37#I<"I<!I< I< c] I< SM I<I<!$I<I<!I<!$I< #I<I< !I<"-1#I<I.AE2Dd3PbKbFc2c,dA@nr''\%(" !%#]^ ]^ " ]^]^r$r'ch^\rSrSrSrU4SjrSS\RS\4Sjjr Sr U=r $) HunyuanDiT2DMultiControlNetModeli8a `HunyuanDiT2DMultiControlNetModel` wrapper class for Multi-HunyuanDiT2DControlNetModel This module is a wrapper for multiple instances of the `HunyuanDiT2DControlNetModel`. The `forward()` API is designed to be compatible with `HunyuanDiT2DControlNetModel`. Args: controlnets (`List[HunyuanDiT2DControlNetModel]`): Provides additional conditioning to the unet during the denoising process. You must set multiple `HunyuanDiT2DControlNetModel` as a list. cX>[TU]5 [R"U5Ulg)N)rHrIrrSnets)r] controlnetsrbs r%rI)HunyuanDiT2DMultiControlNetModel.__init__Es MM+. r$rrc [[X4UR55HQun upnU"UUUUUUUUU U U U S9 nU S:XaUnM$[WSUS5VVs/sH unnUU-PM nnnU4nMS W$s snnf)r) rrrrrrrrrrrrr)rrr)r]rrrrrrrrrrrriimagescaler block_samplescontrol_block_samplescontrol_block_sample block_samples r%r(HunyuanDiT2DMultiControlNetModel.forwardIs\.7s?`d`i`i7j-k )A)j&+! %#(&;$7)A'= /!1' M Av(5%?BBWXYBZ\ijk\l>m)>m:,l)<7>m&)*?(@%1.l4%$ )sA7)rr) rrrr__doc__rIr r!rrr#rrs@r%rr8sP /%(" !%#H% H% " H%H%r$r)! dataclassesrtypingrrrr rconfiguration_utilsr r utilsr r attention_processorr embeddingsrrrmodeling_utilsr#transformers.hunyuan_transformer_2drrrr get_loggerrrrr'rrr$r%rs"(( B(4 (A*   H % 2j2 2L^*kL^^Y%zY%r$