bi ddlZddlmZmZmZmZmZmZddlZ ddl Z ddl Z ddl mZmZddlmZmZddlmZddlmZmZddlmZdd lmZmZmZdd lmZd d l m!Z!d dl"m#Z#er ddl$m%cm&Z'dZ(ndZ(ejRe*Z+dZ, d"de jZdee j\de/fdZ0 d#de1de1de2de2fdZ3 d$dee1deee/e jhfdeee1deee2fdZ5Gd d!e!eZ6y)%N)AnyCallableDictListOptionalUnion)"Qwen2_5_VLForConditionalGenerationQwen2Tokenizer)PipelineImageInputVaeImageProcessor)QwenImageLoraLoaderMixin)AutoencoderKLQwenImageQwenImageTransformer2DModel)FlowMatchEulerDiscreteScheduler)is_torch_xla_availableloggingreplace_example_docstring) randn_tensor)DiffusionPipeline)QwenImagePipelineOutputTFa Examples: ```py >>> import torch >>> from diffusers import QwenImageInpaintPipeline >>> from diffusers.utils import load_image >>> pipe = QwenImageInpaintPipeline.from_pretrained("Qwen/Qwen-Image", torch_dtype=torch.bfloat16) >>> pipe.to("cuda") >>> prompt = "Face of a yellow cat, high resolution, sitting on a park bench" >>> img_url = "https://raw.githubusercontent.com/CompVis/latent-diffusion/main/data/inpainting_examples/overture-creations-5sI6fQgYIuo.png" >>> mask_url = "https://raw.githubusercontent.com/CompVis/latent-diffusion/main/data/inpainting_examples/overture-creations-5sI6fQgYIuo_mask.png" >>> source = load_image(img_url) >>> mask = load_image(mask_url) >>> image = pipe(prompt=prompt, negative_prompt=" ", image=source, mask_image=mask, strength=0.85).images[0] >>> image.save("qwenimage_inpainting.png") ``` encoder_output generator sample_modect|dr |dk(r|jj|St|dr|dk(r|jjSt|dr |jSt d)N latent_distsampleargmaxlatentsz3Could not access latents of provided encoder_output)hasattrrrmoder!AttributeError)rrrs s/home/cdr/jupyterlab/.venv/lib/python3.12/site-packages/diffusers/pipelines/qwenimage/pipeline_qwenimage_inpaint.pyretrieve_latentsr&2st~}-+2I))00;;  /K84K))..00  +%%%RSS base_seq_len max_seq_len base_shift max_shiftc<||z ||z z }|||zz }||z|z}|SN) image_seq_lenr(r)r*r+mbmus r%calculate_shiftr3@s; Z K,$>?AQ%%A  Q B Ir'num_inference_stepsdevice timestepssigmasc | | td|dttj|jj j v}|std|jd|jd ||d||j}t|}||fS|dttj|jj j v}|std|jd|jd ||d||j}t|}||fS|j|fd |i||j}||fS) a Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`. Args: scheduler (`SchedulerMixin`): The scheduler to get timesteps from. num_inference_steps (`int`): The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps` must be `None`. device (`str` or `torch.device`, *optional*): The device to which the timesteps should be moved to. If `None`, the timesteps are not moved. timesteps (`List[int]`, *optional*): Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed, `num_inference_steps` and `sigmas` must be `None`. sigmas (`List[float]`, *optional*): Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed, `num_inference_steps` and `timesteps` must be `None`. Returns: `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the second element is the number of inference steps. zYOnly one of `timesteps` or `sigmas` can be passed. Please choose one to set custom valuesr6zThe current scheduler class zx's `set_timesteps` does not support custom timestep schedules. Please check whether you are using the correct scheduler.)r6r5r7zv's `set_timesteps` does not support custom sigmas schedules. Please check whether you are using the correct scheduler.)r7r5r5r.) ValueErrorsetinspect signature set_timesteps parameterskeys __class__r6len) schedulerr4r5r6r7kwargsaccepts_timesteps accept_sigmass r%retrieve_timestepsrFNs>!3tuu'3w/@/@AXAX/Y/d/d/i/i/k+ll .y/B/B.CDab  M)FMfM'' !)n ) ))   C(9(9):Q:Q(R(](](b(b(d$ee .y/B/B.CD_`  GvfGG'' !)n ) ))  3MFMfM'' ) ))r'c7jeZdZdZdZddgZdededede d e f fd Z d e jd e jfd Z dDdeeeefdee j&dee j(fdZde jde j,fdZdZ dEdeeeefdee j&dedee jdee jdef dZ dFdZedZedZdZd Z d!Z!d"Z" dGd#Z#d$Z$e%d%Z&e%d&Z'e%d'Z(e%d(Z)e%d)Z*e jVe,e-ddd*ddddddd+d,dd-dddddddd.d/dddgd0fdeeeefd1eeeefd2e.de/d3e/d4e/d5eed6eed7eed8e.d9ed:eee.d;e.dedeee j,ee j,fdee jdee jdee jdeed?e0d@ee1ee2fdAee3eee1gdfdBeedef4dCZ4xZ5S)HQwenImageInpaintPipelinea The QwenImage pipeline for text-to-image generation. Args: transformer ([`QwenImageTransformer2DModel`]): Conditional Transformer (MMDiT) architecture to denoise the encoded image latents. scheduler ([`FlowMatchEulerDiscreteScheduler`]): A scheduler to be used in combination with `transformer` to denoise the encoded image latents. vae ([`AutoencoderKL`]): Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations. text_encoder ([`Qwen2.5-VL-7B-Instruct`]): [Qwen2.5-VL-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct), specifically the [Qwen2.5-VL-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct) variant. tokenizer (`QwenTokenizer`): Tokenizer of class [CLIPTokenizer](https://huggingface.co/docs/transformers/en/model_doc/clip#transformers.CLIPTokenizer). ztext_encoder->transformer->vaer! prompt_embedsrBvae text_encoder tokenizer transformerct||j|||||t|ddr"dt |j j znd|_t|ddr |j jjnd|_ t|jdz|j|_ t|jdz|jddd |_ d |_d |_d |_d |_y)N)rJrKrLrMrBrJr)vae_scale_factorvae_latent_channelsFT)rQrR do_normalize do_binarizedo_convert_grayscalez<|im_start|>system Describe the image by detailing the color, shape, size, texture, quantity, text, spatial relationships of the objects and background:<|im_end|> <|im_start|>user {}<|im_end|> <|im_start|>assistant ")super__init__register_modulesgetattrrArJtemperal_downsamplerQconfigz_dimlatent_channelsr image_processormask_processortokenizer_max_lengthprompt_template_encode prompt_template_encode_start_idxdefault_sample_size)selfrBrJrKrLrMr@s r%rZz!QwenImageInpaintPipeline.__init__s  %#  KRRVX]_cJdS)E)E%F Fjk9@eT8Rtxx44XZ0!22Q6DL`L` 0!22Q6 $ 4 4!%  %)!'E#02-#& r' hidden_statesmaskc|j}|jd}||}tj||j d}|S)Nrdimr)boolsumtorchsplittolist)rgrhri bool_mask valid_lengthsselected split_results r%_extract_masked_hiddenz/QwenImageInpaintPipeline._extract_masked_hiddensHIIK ! ! ,  +{{8]-A-A-CK r'Npromptr5dtypec|xs |j}|xs|jj}t|tr|gn|}|j }|j }|Dcgc]}|j|}}|j||j|zdddj|}|j|j|jd} | jd} |j| |j} | Dcgc]}||d } }| Dcgc]A}tj |j#dtj$|j&C} }t)| Dcgc]}|j#dc}} tj*| Dcgc]J}tj,||j/| |j#dz |j#dgLc}}tj*| Dcgc]:}tj,||j/| |j#dz g<c}}|j||}||fScc}wcc}wcc}wcc}wcc}wcc}w) NTpt) max_lengthpadding truncationreturn_tensors) input_idsattention_maskoutput_hidden_statesr)rxr5r)_execution_devicerKrx isinstancestrrdreformatrLrctorrrhrvroonessizelongr5maxstackcat new_zeros)rgrwr5rxtemplatedrop_idxetxt txt_tokensencoder_hidden_statesrhsplit_hidden_statesattn_mask_listr)urIencoder_attention_masks r%_get_qwen_prompt_embedsz0QwenImageInpaintPipeline._get_qwen_prompt_embedss1 14110**00'4&&..88+12axq!22^^ D55@$[_pt$ "V* !% 1 1 **%44!%!2!  .;;B? "99-IbIbc5HIq|II\opWX%**QVVAYejjRpp.AB166!9BC  Vi jQRUYY1;;{QVVAY'>q JK L j "'KY ZaUYY1;;{QVVAY'>?@ A Z" &((uV(D 444/3JpB j [s&I? IAI$"I)AI.=?I3imagerc t|trjt|jdDcgc]1}t |j j |||dz||3}}tj|d}n&t |j j ||}tj|j jjjd|j jjdddj|j|j }dtj|j jj"jd|j jjdddj|j|j z }||z |z}|Scc}w)Nrr)rrk?)rlistrangeshaper&rJencoderortensorr^ latents_meanviewr_rr5rx latents_std)rgrri image_latentsrrs r%_encode_vae_imagez*QwenImageInpaintPipeline._encode_vae_imagesa i &u{{1~.!q1q51A!BiXYl[M"IIm;M,TXX__U-CyYM LL55 6 T!TXX__**Aq! 4 R $$m&9&9 :  ELL)D)DEJJ1dhhooNcNcefhiklmpp  -"5"5  '5D 's6F>cNt||z|}tt||z d}|jj||jj zd}t |jdr2|jj||jj z|||z fS)Nrset_begin_index)minintrrBr6orderr"r)rgr4strengthr5 init_timestept_startr6s r% get_timestepsz&QwenImageInpaintPipeline.get_timesteps s/(:>#4 5 NN * *7T^^5I5I+I J-777r'rnum_images_per_promptprompt_embeds_maskmax_sequence_lengthc|xs |j}t|tr|gn|}| t|n|jd}||j ||\}}|ddd|f}|ddd|f}|j\}} }|j d|d}|j||z| d}|j d|d}|j||z| }||fS)a1 Args: prompt (`str` or `List[str]`, *optional*): prompt to be encoded device: (`torch.device`): torch device num_images_per_prompt (`int`): number of images that should be generated per prompt prompt_embeds (`torch.Tensor`, *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. Nrrr)rrrrArrrepeatr) rgrwr5rrIrr batch_size_seq_lens r% encode_promptz&QwenImageInpaintPipeline.encode_prompts ,1411'4&&$1$9S[}?R?RST?U  040L0LVU[0\ -M-%a)=*=)=&=> /3G4G3G0GH%++ 7A%,,Q0EqI %**:8M+MwXZ[ /66q:OQRS/44ZBW5WY`a000r'c |dks|dkDrtd||jdzzdk7s|jdzzdk7r,tjdjdzd|d|d| Lt fd | Ds8td j d | Dcgc]}|j vs|c}|| td |d | d| | td|7t |ts't |tstdt||| td|d| d| | td| | td|t |tjjstdt|dt |tjjstdt|d|dk7rtd|d||dkDrtd|yycc}w)Nrrz2The value of strength should in [0.0, 1.0] but is rz-`height` and `width` have to be divisible by z but are z and z(. Dimensions will be resized accordinglyc3:K|]}|jvywr-)_callback_tensor_inputs).0krgs r% z8QwenImageInpaintPipeline.check_inputs..[s#F 23A-- -F sz2`callback_on_step_end_tensor_inputs` has to be in z , but found zCannot forward both `prompt`: z and `prompt_embeds`: z2. Please make sure to only forward one of the two.zeProvide either `prompt` or `prompt_embeds`. Cannot leave both `prompt` and `prompt_embeds` undefined.z2`prompt` has to be of type `str` or `list` but is z'Cannot forward both `negative_prompt`: z and `negative_prompt_embeds`: zIf `prompt_embeds` are provided, `prompt_embeds_mask` also have to be passed. Make sure to generate `prompt_embeds_mask` from the same text encoder that was used to generate `prompt_embeds`.zIf `negative_prompt_embeds` are provided, `negative_prompt_embeds_mask` also have to be passed. Make sure to generate `negative_prompt_embeds_mask` from the same text encoder that was used to generate `negative_prompt_embeds`.zJThe image should be a PIL image when inpainting mask crop, but is of type .zOThe mask image should be a PIL image when inpainting mask crop, but is of type pilz@The output type should be PIL when inpainting mask crop, but is rVz9`max_sequence_length` cannot be greater than 1024 but is ) r9rQloggerwarningallrrrrtypePILImage)rgrwr mask_imagerheightwidth output_typenegative_promptrInegative_prompt_embedsrnegative_prompt_embeds_mask"callback_on_step_end_tensor_inputspadding_mask_croprrs` r% check_inputsz%QwenImageInpaintPipeline.check_inputsAs$ a<8a<QRZQ[\] ] T**Q. /1 4AVAVYZAZ8[_`8` NN?@U@UXY@Y?ZZcdjckkpqvpwx`a  . 9#F 7YF C DTEaEaDbbn|^pHvwbckokGkGbGpqpHoIJ   -";08N}o^00 ^ 5w  FC)@TZ\`IaQRVW]R^Q_`a a  &+A+M9/9JK*++]_   $);)CQ  " -2M2Uu   (eSYY__5 `aefkal`mmnoj#))//: Z(),e# #cdocppq!rss  */BT/IXYlXmno o0J *UpHs G51G5c|j|||dzd|dzd}|jdddddd}|j||dz|dzz|dz}|S)Nrrrr )rpermutereshape)r!rnum_channels_latentsrrs r% _pack_latentsz&QwenImageInpaintPipeline._pack_latentssl,,z+?1aQVZ[Q[]^_//!Q1a3//*v{uz.JL`cdLder'c|j\}}}dt||dzzz}dt||dzzz}|j||dz|dz|dzdd}|jdddddd}|j ||dzd||}|S)Nrrrr rr)rrrrr)r!rrrQr num_patcheschannelss r%_unpack_latentsz(QwenImageInpaintPipeline._unpack_latentss-4MM) Kc&k&6&:;<SZ$4q$89:,,z6Q; HPQMSTVWX//!Q1a3//*h5.A1feTr'c8|jjy)z Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to compute decoding in several steps. This is useful to save some memory and allow larger batch sizes. N)rJenable_slicingrgs r%enable_vae_slicingz+QwenImageInpaintPipeline.enable_vae_slicing !r'c8|jjy)z Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to computing decoding in one step. N)rJdisable_slicingrs r%disable_vae_slicingz,QwenImageInpaintPipeline.disable_vae_slicings   "r'c8|jjy)a Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow processing larger images. N)rJ enable_tilingrs r%enable_vae_tilingz*QwenImageInpaintPipeline.enable_vae_tilings  r'c8|jjy)z Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to computing decoding in one step. N)rJdisable_tilingrs r%disable_vae_tilingz+QwenImageInpaintPipeline.disable_vae_tilingrr'c t| tr)t| |k7rtdt| d|ddt ||j dzzz}dt ||j dzzz}|d|||f} |j dk(r|jd}n0|j dk7rtd|j d | | j|| S|j|| }|jd|jk7r|j|| } n|} || jd kDrC|| jd zd k(r.|| jd z} tj| g| zd } n^|| jd kDr4|| jd zd k7rtd| jd d|dtj| gd } | jdd} | -t| | ||}|j j#| ||} n| j|}|} |j%|||||}|j%| ||||} |j%| ||||} | || fS)Nz/You have passed a list of generators of length z+, but requested an effective batch size of z@. Make sure the batch size matches the length of the generators.rrrr Expected image dims 4 or 5, got rr5rxrrrrkz'Cannot duplicate `image` of batch size z to z text prompts.)rr5rx)rrrAr9rrQrl unsqueezerrr`rror transposerrB scale_noiser)rgrtimesteprrrrrxr5rr!rradditional_image_per_promptnoises r%prepare_latentsz(QwenImageInpaintPipeline.prepare_latentss i &3y>Z+GA#i.AQR&<'gi  c&kd&;&;a&?@ASZD$9$9A$=>?Q 4feD 99;! OOA&E YY[A ? }ANO O  ::V5:9 9e4 ;;q>T11 1 22)2TM!M  ++A. .: @S@STU@V3VZ[3[*4 8K8KA8N*N '!II}o8S&SYZ[M ---a0 0Z-BUBUVWBX5X\]5]9-:M:Ma:P9QQUV`Uaaop "II}o1=M%//15 ? )FRWXEnn00%PGJJv&EG""5*6JFTYZ**=*FZ\bdij $$Wj:NPVX]^},,r'c dt||jdzzz}dt||jdzzz}tjjj |||f}|j | |}||z}|jdk(r|jd}n0|jdk7rtd|jd|j | |}|jd|jk(r|} n|j|| } |jd |krX||jd zd k(std |d |jd d |j||jd zddd}| jd |krY|| jd zd k(std|d | jd d| j|| jd zdddd} | j | |} |j| ||||} |j|jd|dd||||}|| fS)Nr)rrrrrrrrrzvThe passed mask and the required batch size don't match. Masks are supposed to be duplicated to a total batch size of z, but zo masks were passed. Make sure the number of masks that you pass is divisible by the total requested batch size.zyThe passed images and the required batch size don't match. Images are supposed to be duplicated to a total batch size of zq images were passed. Make sure the number of images that you pass is divisible by the total requested batch size.)rrQronn functional interpolaterrlrr9rr`rrr) rgri masked_imagerrrrrrxr5rmasked_image_latentss r%prepare_mask_latentsz-QwenImageInpaintPipeline.prepare_mask_latentssc&kd&;&;a&?@ASZD$9$9A$=>?xx""..t65/.JwwfEw2"77     "'11!4L    1 $? @P@P@R?SSTUV V#fEB   a D$8$8 8#/ #'#9#9 Xa#9#b  ::a=: % 1 -2 ..8\ 1 O^^ ;;zTZZ]:Aq!DD  % %a (: 5 4 : :1 ==B 11; FCWC]C]^_C`Babtt $8#>#>zMaMgMghiMj?jlmoprsuv#w  466fE6R#11     !! KK/A 6     )))r'c|jSr-)_guidance_scalers r%guidance_scalez'QwenImageInpaintPipeline.guidance_scaleOs###r'c|jSr-)_attention_kwargsrs r%attention_kwargsz)QwenImageInpaintPipeline.attention_kwargsS%%%r'c|jSr-)_num_timestepsrs r% num_timestepsz&QwenImageInpaintPipeline.num_timestepsWs"""r'c|jSr-)_current_timesteprs r%current_timestepz)QwenImageInpaintPipeline.current_timestep[rr'c|jSr-) _interruptrs r% interruptz"QwenImageInpaintPipeline.interrupt_s r'g@g333333?2rrTirtrue_cfg_scalerrrrrrr4r7rrrr return_dictrcallback_on_step_endrc\|xs|j|jz}|xs|j|jz}|j|||| |||||||||| || |_||_d|_d|_| "|jj|||| }d}nd}d}|}|jj|||||}|jtj}|t|trd }n-|t|t r t#|}n|j$d }|j&} |duxs |duxr|du}!|d kDxr|!}"|j)|||| || \}}|"r|j)|||| || \}}| t+j,d d | z | n| } t/||jzd zt/||jzd zz}#t1|#|j2j4j7dd|j2j4j7dd|j2j4j7dd|j2j4j7dd}$t9|j2| | | |$\}%} |j;| | | \}%} | d krt=d| d| d|%dd j?||z}&|j@j4jBdz}'|jE||&||z|'|||jF| || \}}(})|jj|||||}*| ||*dkz}+n|}+|jI|*|+||'||||jF| | \},}d ||jzd z||jzd zfgg|z}-tKt#|%| |j2jLzz d }.t#|%|_'|j@j4jPrGtjRd g| | tj}/|/jU|j$d }/nd}/|jVi|_| |jYd j[nd}0| |jYd j[nd}1|j]| 5}2t_|%D]\}3}4|j`r|4|_|4jU|j$d j|jF}5|j@jcd5|jA||5d z |/|||-|0|jVd! d }6ddd|"r|j@jcd"5|jA||5d z |/|||-|1|jVd! d }7ddd7|6|7z zz}8tjd|6d#d$%}9tjd|8d#d$%}:|8|9|:z z}6|jF};|j2jg6|4|d&d }|)}<|,}=|3t#|%d z kr9|%|3d z}>|j2ji|g|(} 1.0 and a provided `negative_prompt`, enables true classifier-free guidance. mask_image (`torch.Tensor`, `PIL.Image.Image`, `np.ndarray`, `List[torch.Tensor]`, `List[PIL.Image.Image]`, or `List[np.ndarray]`): `Image`, numpy array or tensor representing an image batch to mask `image`. White pixels in the mask are repainted while black pixels are preserved. If `mask_image` is a PIL image, it is converted to a single channel (luminance) before use. If it's a numpy array or pytorch tensor, it should contain one color channel (L) instead of 3, so the expected shape for pytorch tensor would be `(B, 1, H, W)`, `(B, H, W)`, `(1, H, W)`, `(H, W)`. And for numpy array would be for `(B, H, W, 1)`, `(B, H, W)`, `(H, W, 1)`, or `(H, W)`. mask_image_latent (`torch.Tensor`, `List[torch.Tensor]`): `Tensor` representing an image batch to mask `image` generated by VAE. If not provided, the mask latents tensor will ge generated by `mask_image`. height (`int`, *optional*, defaults to self.unet.config.sample_size * self.vae_scale_factor): The height in pixels of the generated image. This is set to 1024 by default for the best results. width (`int`, *optional*, defaults to self.unet.config.sample_size * self.vae_scale_factor): The width in pixels of the generated image. This is set to 1024 by default for the best results. padding_mask_crop (`int`, *optional*, defaults to `None`): The size of margin in the crop to be applied to the image and masking. If `None`, no crop is applied to image and mask_image. If `padding_mask_crop` is not `None`, it will first find a rectangular region with the same aspect ration of the image and contains all masked area, and then expand that area based on `padding_mask_crop`. The image and mask_image will then be cropped based on the expanded area before resizing to the original image size for inpainting. This is useful when the masked area is small while the image is large and contain information irrelevant for inpainting, such as background. strength (`float`, *optional*, defaults to 1.0): Indicates extent to transform the reference `image`. Must be between 0 and 1. `image` is used as a starting point and more noise is added the higher the `strength`. The number of denoising steps depends on the amount of noise initially added. When `strength` is 1, added noise is maximum and the denoising process runs for the full number of iterations specified in `num_inference_steps`. A value of 1 essentially ignores `image`. num_inference_steps (`int`, *optional*, defaults to 50): The number of denoising steps. More denoising steps usually lead to a higher quality image at the expense of slower inference. sigmas (`List[float]`, *optional*): Custom sigmas to use for the denoising process with schedulers which support a `sigmas` argument in their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is passed will be used. guidance_scale (`float`, *optional*, defaults to 3.5): Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://huggingface.co/papers/2207.12598). `guidance_scale` is defined as `w` of equation 2. of [Imagen Paper](https://huggingface.co/papers/2205.11487). Guidance scale is enabled by setting `guidance_scale > 1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`, usually at the expense of lower image quality. This parameter in the pipeline is there to support future guidance-distilled models when they come up. Note that passing `guidance_scale` to the pipeline is ineffective. To enable classifier-free guidance, please pass `true_cfg_scale` and `negative_prompt` (even an empty negative prompt like " ") should enable classifier-free guidance computations. 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.Tensor`, *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 (`torch.Tensor`, *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 (`torch.Tensor`, *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.qwenimage.QwenImagePipelineOutput`] instead of a plain tuple. 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` defaults to 512): Maximum sequence length to use with the `prompt`. Examples: Returns: [`~pipelines.qwenimage.QwenImagePipelineOutput`] or `tuple`: [`~pipelines.qwenimage.QwenImagePipelineOutput`] if `return_dict` is True, otherwise a `tuple`. When returning a tuple, the first element is a list with the generated images. ) rrrIrrrrrrNF)padfilldefault)rr crops_coords resize_mode)rxrr)rwrIrr5rrrrbase_image_seq_lenmax_image_seq_lenr*?r+ffffff?)r7r2z?After adjusting the num_inference_steps by strength parameter: z!, the number of pipelinesteps is z4 which is < 1 and not appropriate for this pipeline.r)rrrrrrk)totalcondi) rhrguidanceencoder_hidden_states_maskr img_shapes txt_seq_lensrruncondrT)rlkeepdim)rr!rIlatent)r)images)KrfrQrrrrr rbget_crop_regionra preprocessrrofloat32rrrrArrrnplinspacerr3rBr^getrFrr9rrM in_channelsrrxrrrrguidance_embedsfullexpandrrnrq progress_bar enumerater  cache_contextnormsteprrbackendsmps is_availablelocalspopupdate XLA_AVAILABLExm mark_steprrJrrr_r5rdecode postprocess apply_overlaymaybe_free_model_hooksr)Drgrwrrrrrrrrrr4r7rrrr!rIrrrrrrrrrrroriginal_image init_imagerr5has_neg_prompt do_true_cfgr/r2r6latent_timesteprrrmask_conditionrrir!num_warmup_stepsrr"negative_txt_seq_lensr1rtr noise_predneg_noise_pred comb_pred cond_norm noise_norm latents_dtypeinit_latents_proper init_masknoise_timestepcallback_kwargsrcallback_outputsrrsD r%__call__z!QwenImageInpaintPipeline.__call__cs@ VK433d6K6KKI11D4I4II       #+'#91(C/Q/ 3  $ .!1!%  (..>>z5RX^o>pL KL#K))44 &LVa5  ]]]7   *VS"9J  Jvt$<VJ&,,Q/J''(4 "$ . Z3NVZ3Z %q(;^ ,0,>,>'1"7 3 -?- ) ) BFBTBT&4#>&;$7 CUC ? "$?TZSaS!&9"9;NOgmV(=(==Bs5zUYUjUjGjnoGop   NN ! ! % %&:C @ NN ! ! % %&94 @ NN ! ! % %lC 8 NN ! ! % %k4 8  *< NN   * & &*.););L/L%)%>%>    !        & ""6T%:%::a?$J_J_A_cdAdefgjtt s9~0CdnnFZFZ0ZZ\]^!)n    " " 2 2zz1#~fEMMZHw}}Q'78HH  (%'D "ASA_)--!-4;;=ei ?Z?f ' + + + 2 9 9 ;lp    %8  9K #\!),J #1>>)*&88GMM!$4588G%%33F; !%!1!1&-!)D!)3E.;#-%1)-)>)>$)"2 " "J ))77A )-)9)9*1%-_%-7R2H'1)>-1-B-B(-*: * * !/:P^C^1_ _I % :2t LI!&I2t!LJ!*i*.D!EJ!( ..--j!WRW-XYZ['4# s9~))%.q1u%5N*...*D*D+U\\>:J-KU+'y=,??)gBUU==M1~~))668")**]";'3&(O?9-3Xa[*9';D!Q'X$.229gFG$4$8$8-$XMI**A9I/IqSTuX\XfXfXlXlNlpqNq '') LLNUJ #K #Z"& ( "E**7FE4CXCXYGjj0G TXX__99:a..1a8GNNGMM2   TXX__-H-H I N NqRVRZRZRaRaRgRgijlmop q t t !K +l:GHHOOGO?B1a7KE((44U 4TE ,mrhiD((66z>STVbc ##%8O&e44G    -K #K #BsJ4A9f-+f%f=+f(F5ff:%f)f ff ff&)NNN)NrNNrV)NNNNNNNNr-)6__name__ __module__ __qualname____doc__model_cpu_offload_seqrrrr r rrZroTensorrvrrrrr5rxr Generatorrrrrr staticmethodrrrrrrrrpropertyrrrr r no_gradrEXAMPLE_DOC_STRINGfloatr rmrrrrW __classcell__)r@s@r%rHrHsb$=(/:"'2"'$"'9 "' " "' 1 "'JELL )-)-'+ $5c49n%$5&$5 $ $5Nu||0 8*.%&0459#''1c49n%'1&'1 # '1   - '1 %U\\2 '1!'1d#$(+/ !IpV  "#!"$?-BI*V$$&&##&&U]]_12)-15 #$()-37 $#+/#%(, #%&MQ*.04599=>B%* 59KO9B #&7R5c49n%R5sDI~.R5 R5 " R5 ' R51R5 R5}R5$C=R5R5!R5e%R5R5 #R5 E%//43H"HIJ!R5"%,,'#R5$  -%R5&%U\\2'R5(!) 6)R5*&.ell%;+R5,c]-R5./R50#4S>21R52'xc40@$0F'GH3R54-1I5R56!7R53R5r'rH)Nr)rrrr)NNNN)7r;typingrrrrrrnumpyr* PIL.Imagerro transformersr r rar r loadersrmodelsrr schedulersrutilsrrrutils.torch_utilsrpipeline_utilsrpipeline_outputrtorch_xla.core.xla_modelcore xla_modelr=r< get_loggerrXrrbr]r^rr&rrcr3r5rFrHr.r'r%rtsL== KD/I9OO-.4))MM   H %,ck TLL T-5eoo-F T\_ T        *.15%)$( 8*!#8* U3 ,- .8*S " 8* T%[ ! 8*vn 502Jn 5r'