3jg tSSKrSSKJrJr SSKrSSKJrJr SSKJ r SSK J r J r SSK Jr SSKJr SS KJr SS KJr SS KJrJr SS KJr S SKJr \R8"\5rSrSS\ S\ S\!S\!4Sjjr"SS\ S-S\#\RH-S-S\%\ S-S\%\!S-4Sjjr&"SS\\ \ 5r'g)N)AnyCallable) AutoTokenizerPreTrainedModel)VaeImageProcessor)FromSingleFileMixinZImageLoraLoaderMixin) AutoencoderKL)ZImageTransformer2DModel)DiffusionPipeline)FlowMatchEulerDiscreteScheduler)loggingreplace_example_docstring) randn_tensor)ZImagePipelineOutputut Examples: ```py >>> import torch >>> from diffusers import ZImagePipeline >>> pipe = ZImagePipeline.from_pretrained("Z-a-o/Z-Image-Turbo", torch_dtype=torch.bfloat16) >>> pipe.to("cuda") >>> # Optionally, set the attention backend to flash-attn 2 or 3, default is SDPA in PyTorch. >>> # (1) Use flash attention 2 >>> # pipe.transformer.set_attention_backend("flash") >>> # (2) Use flash attention 3 >>> # pipe.transformer.set_attention_backend("_flash_3") >>> prompt = "一幅为名为“造相「Z-IMAGE-TURBO」”的项目设计的创意海报。画面巧妙地将文字概念视觉化：一辆复古蒸汽小火车化身为巨大的拉链头，正拉开厚厚的冬日积雪，展露出一个生机盎然的春天。" >>> image = pipe( ... prompt, ... height=1024, ... width=1024, ... num_inference_steps=9, ... guidance_scale=0.0, ... generator=torch.Generator("cuda").manual_seed(42), ... ).images[0] >>> image.save("zimage.png") ``` base_seq_len max_seq_len base_shift max_shiftc4XC- X!- - nX5U-- nX-U-nU$N) image_seq_lenrrrrmbmus f/home/wildlama/miniconda3/lib/python3.13/site-packages/diffusers/pipelines/z_image/pipeline_z_image.pycalculate_shiftr @s3  K$>?A%%A  Q B Inum_inference_stepsdevice timestepssigmasc UbUb [S5eUbS[[R"UR5R R 55;nU(d[SURS35eUR"S X2S.UD6 URn[U5nX14$UbS[[R"UR5R R 55;nU(d[SURS35eUR"S XBS.UD6 URn[U5nX14$UR"U4S U0UD6 URnX14$) a Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`. Args: scheduler (`SchedulerMixin`): The scheduler to get timesteps from. num_inference_steps (`int`): The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps` must be `None`. device (`str` or `torch.device`, *optional*): The device to which the timesteps should be moved to. If `None`, the timesteps are not moved. timesteps (`list[int]`, *optional*): Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed, `num_inference_steps` and `sigmas` must be `None`. sigmas (`list[float]`, *optional*): Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed, `num_inference_steps` and `timesteps` must be `None`. Returns: `tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the second element is the number of inference steps. zYOnly one of `timesteps` or `sigmas` can be passed. Please choose one to set custom valuesr$zThe current scheduler class zx's `set_timesteps` does not support custom timestep schedules. Please check whether you are using the correct scheduler.)r$r#r%zv's `set_timesteps` does not support custom sigmas schedules. Please check whether you are using the correct scheduler.)r%r#r#r) ValueErrorsetinspect signature set_timesteps parameterskeys __class__r$len) schedulerr"r#r$r%kwargsaccepts_timesteps accept_sigmass rretrieve_timestepsr4Ns}>!3tuu'3w/@/@AXAX/Y/d/d/i/i/k+ll .y/B/B.CDab  M)MfM'' !)n  ))   C(9(9):Q:Q(R(](](b(b(d$ee .y/B/B.CD_`  GvGG'' !)n  ))  3MFMfM''  ))r!c+^\rSrSrSr/rSS/rS\S\S\ S\ S \ 4 U4S jjr S3S\ \\ -S\R S -S\S\ \\ -S -S\\R$S -S\R$S -S\4SjjrS4S\ \\ -S\R S -S\\R$S -S\S\\R$4 SjjrS5Sjr\S5r\S5r\S5r\S5r\S5r\R:"5\"\5S S S SS SSS S S!S S S S S"S S S S/S 4S\ \\ -S#\S -S$\S -S%\S&\\ S -S'\ S(\S)\ S\ \\ -S -S*\S -S+\RB\\RB-S -S\R$S -S\\R$S -S\\R$S -S,\ S -S-\S.\"\ \#4S -S/\$\\/S 4S -S0\\ S\4(S1jj55r%S2r&U=r'$)6ZImagePipelineztext_encoder->transformer->vaelatents prompt_embedsr0vae text_encoder tokenizer transformerc&>[TU]5 URUUUUUS9 [US5(a<URb/S[ URR R5S- -OSUl[URS-S9Ul g)N)r:r;r<r0r=r:r)vae_scale_factor) super__init__register_moduleshasattrr:r/configblock_out_channelsrArimage_processor)selfr0r:r;r<r=r.s rrCZImagePipeline.__init__s  %#  CJ$PUBVBV[_[c[c[oA#dhhoo889A= >uv  1$BWBWZ[B[\r!NTpromptr#do_classifier_free_guidancenegative_promptnegative_prompt_embedsmax_sequence_lengthc:[U[5(aU/OUnURUUUUS9nU(a_UcUVs/sHnSPM nnO[U[5(aU/OUn[U5[U5:XdeURUUUUS9nXV4$/nXV4$s snf)N)rLr#r9rP) isinstancestr_encode_promptr/) rIrLr#rMrNr9rOrP_s r encode_promptZImagePipeline.encode_prompts(44&&++' 3 ,  '&/5"6v!2v"67A/SV7W7W?"3]lv;#o"66 66%)%8%8&4$7 &9& "44&( "44#7s Breturnc2U=(d URnUbU$[U[5(aU/n[U5H+upVSUS./nURR USSSS9nXaU'M- UR USUSSS9nUR RU5n URRU5R5n URU U SS 9RS n/n [[U55HnU RX5X5 M U $) Nuser)rolecontentFT)tokenizeadd_generation_promptenable_thinking max_lengthpt)paddingra truncationreturn_tensors) input_idsattention_maskoutput_hidden_states)_execution_devicerSrT enumerater<apply_chat_templaterftorgboolr; hidden_statesranger/append) rIrLr#r9rPi prompt_itemmessages text_inputstext_input_ids prompt_masksembeddings_lists rrUZImagePipeline._encode_promptsF1411  $ fc " "XF'/NAK8H..<<&* $ =K $1I0nn  * % %..11&9"1144V<AAC ))$'!%*  -  s=)*A  " "=#3LO#D E+r!c S[U5URS---nS[U5URS---nXX44n Uc [XXeS9nU$URU :wa[ SURSU 35eUR U5nU$)Nr?) generatorr#dtypezUnexpected latents shape, got z , expected )intrArshaper'rm) rI batch_sizenum_channels_latentsheightwidthr|r#r{r8r~s rprepare_latentsZImagePipeline.prepare_latentssc&kd&;&;a&?@ASZD$9$9A$=>?6A ?"5fZG }}% #A'--P[\a[b!cddjj(Gr!cUR$r_guidance_scalerIs rguidance_scaleZImagePipeline.guidance_scales###r!c URS:$)Nrrrs rrM*ZImagePipeline.do_classifier_free_guidances##a''r!cUR$r)_joint_attention_kwargsrs rjoint_attention_kwargs%ZImagePipeline.joint_attention_kwargss+++r!cUR$r)_num_timestepsrs r num_timestepsZImagePipeline.num_timestepss"""r!cUR$r) _interruptrs r interruptZImagePipeline.interrupt!s r!2g@Fg?rpilrrr"r%rcfg_normalizationcfg_truncationnum_images_per_promptr{ output_type return_dictrcallback_on_step_end"callback_on_step_end_tensor_inputsc VU=(d SnU=(d SnURS-nUU-S:wa[SUSUSUS35eUU-S:wa[SUSUS US35eURnX`lUUlS UlXplXlUb[U[5(aS nO/Ub![U[5(a [U5nO [U 5nU b#Uc UR(aUc [S 5eO"URUU URU UUUS9un nURRnUR!UU -UUU["R$UU U 5n U S :afU VVs/sHn['U 5HnUPM M n nnUR(a.U(a'UVVs/sHn['U 5HnUPM M nnnUU -nU R(SS-U R(SS--n[+UUR,R.R1SS5UR,R.R1SS5UR,R.R1SS5UR,R.R1SS55nSUR,lSU0n[5UR,UU4SU0UD6un n[7[U 5X@R,R8-- S5n![U 5UlUR,R=S5 UR(aKURb>[?UR5S ::a%SU R?5- S- RA5n"OS n"URCUS9n#[EU 5GHun$n%URF(aMU%RIU R(S5n&SU&- S- n&URJn'U"bU"U$UR:aSn'UR=(a U'S:n(U((aOU RMURRN5n)U)RQSS S S 5n*X-n+U&RQS5n,O)U RMURRN5n*U n+U&n,U*RSS5n*[U*RUSS95n-URU-U,U+S S9Sn.U((GaU.S Un/U.US n0/n1['U5Hn2U/U2R?5n3U0U2R?5n4U3U'U3U4- --n5UR (a}[?UR 5S:ad["RVRYU35n6["RVRYU55n7U6[?UR 5-n8U7U8:aU5U8U7- -n5U1R[U55 M ["R\"U1SS9n1O4["R\"U.V%s/sHn%U%R?5PM sn%SS9n1U1R_S5n1U1*n1UR,RaU1RM["R$5W%U S S9Sn U RN["R$:XdeUb\0n9UHn:[c5U:U9U:'M U"UU$U%U95n;U;ReSU 5n U;ReS U 5n U;ReS!U5nU$[U 5S - :Xd/U$S -U!:dGMfU$S -UR,R8-S:XdGMU#Rg5 GM S S S 5 US":XaU n 1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`, usually at the expense of lower image quality. cfg_normalization (`bool`, *optional*, defaults to False): Whether to apply configuration normalization. cfg_truncation (`float`, *optional*, defaults to 1.0): The truncation value for configuration. negative_prompt (`str` or `list[str]`, *optional*): The prompt or prompts not to guide the image generation. If not defined, one has to pass `negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is less than `1`). num_images_per_prompt (`int`, *optional*, defaults to 1): The number of images to generate per prompt. generator (`torch.Generator` or `list[torch.Generator]`, *optional*): One or a list of [torch generator(s)](https://pytorch.org/docs/stable/generated/torch.Generator.html) to make generation deterministic. latents (`torch.FloatTensor`, *optional*): Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image generation. Can be used to tweak the same generation with different prompts. If not provided, a latents tensor will be generated by sampling using the supplied random `generator`. prompt_embeds (`list[torch.FloatTensor]`, *optional*): Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not provided, text embeddings will be generated from `prompt` input argument. negative_prompt_embeds (`list[torch.FloatTensor]`, *optional*): Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input argument. output_type (`str`, *optional*, defaults to `"pil"`): The output format of the generate image. Choose between [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`. return_dict (`bool`, *optional*, defaults to `True`): Whether or not to return a [`~pipelines.stable_diffusion.ZImagePipelineOutput`] instead of a plain tuple. joint_attention_kwargs (`dict`, *optional*): A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under `self.processor` in [diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py). callback_on_step_end (`Callable`, *optional*): A function that calls at the end of each denoising steps during the inference. The function is called with the following arguments: `callback_on_step_end(self: DiffusionPipeline, step: int, timestep: int, callback_kwargs: Dict)`. `callback_kwargs` will include a list of all tensors as specified by `callback_on_step_end_tensor_inputs`. callback_on_step_end_tensor_inputs (`List`, *optional*): The list of tensor inputs for the `callback_on_step_end` function. The tensors specified in the list will be passed as `callback_kwargs` argument. You will only be able to include variables listed in the `._callback_tensor_inputs` attribute of your pipeline class. max_sequence_length (`int`, *optional*, defaults to 512): Maximum sequence length to use with the `prompt`. Examples: Returns: [`~pipelines.z_image.ZImagePipelineOutput`] or `tuple`: [`~pipelines.z_image.ZImagePipelineOutput`] if `return_dict` is True, otherwise a `tuple`. When returning a tuple, the first element is a list with the generated images. ir?rzHeight must be divisible by z (got z-). Please adjust the height to a multiple of .zWidth must be divisible by z,). Please adjust the width to a multiple of FNrzWhen `prompt_embeds` is provided without `prompt`, `negative_prompt_embeds` must also be provided for classifier-free guidance.)rLrNrMr9rOr#rPrbase_image_seq_lenmax_image_seq_lenr?rffffff?grr%i)total)dim)rr8r9rOlatent)r)images)> q)Q.7==3Cq3HI   NN ! ! % %&:C @ NN ! ! % %&94 @ NN ! ! % %lC 8 NN ! ! % %k4 8  $'  ":); NN  *  *  * & &s9~0CnnFZFZ0ZZ\]^!)n &&q)  + +0D0D0PUZ[_[o[oUptuUu%)IOO,=%=$E#M#M#O #'   %8  9\!),1>>88GMM!$45 8Ot3)-)<)<&'3+A.1E1EE14.!<<[AWZ[A[ $+JJt/?/?/E/E$FM)6)=)=aAq)I&0=0V-+3??1+=()0D4D4D4J4J)K&0=-+3(%7%A%A!%D"*./A/H/HQ/H/O*P'!%!1!1+-AC\jo"2"",-?.?@G,->-?@G!#J"#45%aj..0%aj..0"%;sSy%II 22uT=T=T7UX[7[+0<<+C+CC+HL+0<<+C+CD+IL+7%@W@W:X+XL+l:'+|l/J'K"))$/6 "'ZQ!?J!&-PAaggi-PVW!XJ'//2 ([ ..--jmmEMM.JAwdi-jklm}} 555'3&(O?-3Xa[*@';D!Q'X$.229gFG$4$8$8-$XM-=-A-ABZ\r-s*I**A9I/IqSTuX\XfXfXlXlNlpqNq '')[-:` ( "Ejj0G!?!??488??C_C__GHHOOGO?BE((44U 4TE ##%8O#511K_)vx.Qs: 9s8*` )` I%`1` C.`>`!`` `()rrrrrrrHrA)NTNNNrK)NNrKr)(__name__ __module__ __qualname____firstlineno__model_cpu_offload_seq_optional_components_callback_tensor_inputsrr rrr rCrTrrr#rn FloatTensorr}rWrUrpropertyrrMrrrno_gradrEXAMPLE_DOC_STRINGr Generatordictrrr__static_attributes__ __classcell__)r.s@rr6r6s<(/:]2]]& ] ! ] . ]2'+,0268<;?#& 5d3i 5 t# 5&* 5 tCy4/ 5 E--.5 5!& 1 1D 8 5! 5J'+8<#& 1d3i1 t#1E--.5 1 ! 1 e 1x0$$((,,## ]]_12#'! #%%) #"' #26,-DH,08l2 4Z!l2"#l2$!%S#X 5%l2&'Sz4'784?'l2(-1I)l2*!+l23l2r!r6)rrrr)NNNN)(r)typingrrr transformersrrrHrloadersr r models.autoencodersr models.transformersr pipelines.pipeline_utilsr schedulersrutilsrrutils.torch_utilsrpipeline_outputr get_loggerrloggerrr}rr rTr#rr4r6rr!rr s 70A0;997-1   H %@       '+(,"&!% 8*t8* %,,  %8*Cy4 8* K$  8*vJ2&(=?RJ2r!