src_inference/pipeline.py

import inspect
from typing import Any, Callable, Dict, List, Optional, Union

import numpy as np
import torch
from transformers import CLIPTextModel, CLIPTokenizer, T5EncoderModel, T5TokenizerFast

from diffusers.image_processor import (VaeImageProcessor)
from diffusers.loaders import FluxLoraLoaderMixin, FromSingleFileMixin
from diffusers.models.autoencoders import AutoencoderKL
from diffusers.schedulers import FlowMatchEulerDiscreteScheduler
from diffusers.utils import (
    USE_PEFT_BACKEND,
    is_torch_xla_available,
    logging,
    scale_lora_layers,
    unscale_lora_layers,
)
from diffusers.utils.torch_utils import randn_tensor
from diffusers.pipelines.pipeline_utils import DiffusionPipeline
from diffusers.pipelines.flux.pipeline_output import FluxPipelineOutput
from torchvision.transforms.functional import pad
from diffusers import FluxTransformer2DModel

if is_torch_xla_available():
    import torch_xla.core.xla_model as xm

    XLA_AVAILABLE = True
else:
    XLA_AVAILABLE = False

logger = logging.get_logger(__name__)  # pylint: disable=invalid-name

def calculate_shift(
        image_seq_len,
        base_seq_len: int = 256,
        max_seq_len: int = 4096,
        base_shift: float = 0.5,
        max_shift: float = 1.16,
):
    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
    b = base_shift - m * base_seq_len
    mu = image_seq_len * m + b
    return mu

def prepare_latent_image_ids_(height, width, device, dtype):
    latent_image_ids = torch.zeros(height//2, width//2, 3, device=device, dtype=dtype)
    latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height//2, device=device)[:, None]  # y
    latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width//2, device=device)[None, :]   # x
    return latent_image_ids

def prepare_latent_subject_ids(height, width, device, dtype):
    latent_image_ids = torch.zeros(height // 2, width // 2, 3, device=device, dtype=dtype)
    latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height // 2, device=device)[:, None]
    latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width // 2, device=device)[None, :]
    latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
    latent_image_ids = latent_image_ids.reshape(
        latent_image_id_height * latent_image_id_width, latent_image_id_channels
    )
    return latent_image_ids.to(device=device, dtype=dtype)

def resize_position_encoding(batch_size, original_height, original_width, target_height, target_width, device, dtype):
    latent_image_ids = prepare_latent_image_ids_(original_height, original_width, device, dtype)
    latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
    latent_image_ids = latent_image_ids.reshape(
        latent_image_id_height * latent_image_id_width, latent_image_id_channels
    )
    
    scale_h = original_height / target_height
    scale_w = original_width / target_width
    latent_image_ids_resized = torch.zeros(target_height//2, target_width//2, 3, device=device, dtype=dtype)
    latent_image_ids_resized[..., 1] = latent_image_ids_resized[..., 1] + torch.arange(target_height//2, device=device)[:, None] * scale_h
    latent_image_ids_resized[..., 2] = latent_image_ids_resized[..., 2] + torch.arange(target_width//2, device=device)[None, :] * scale_w
    
    cond_latent_image_id_height, cond_latent_image_id_width, cond_latent_image_id_channels = latent_image_ids_resized.shape
    cond_latent_image_ids = latent_image_ids_resized.reshape(
            cond_latent_image_id_height * cond_latent_image_id_width, cond_latent_image_id_channels
        )
    return latent_image_ids, cond_latent_image_ids 
    
# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
def retrieve_latents(
        encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
):
    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
        return encoder_output.latent_dist.sample(generator)
    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
        return encoder_output.latent_dist.mode()
    elif hasattr(encoder_output, "latents"):
        return encoder_output.latents
    else:
        raise AttributeError("Could not access latents of provided encoder_output")


# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
def retrieve_timesteps(
        scheduler,
        num_inference_steps: Optional[int] = None,
        device: Optional[Union[str, torch.device]] = None,
        timesteps: Optional[List[int]] = None,
        sigmas: Optional[List[float]] = None,
        **kwargs,
):
    if timesteps is not None and sigmas is not None:
        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
    if timesteps is not None:
        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
        if not accepts_timesteps:
            raise ValueError(
                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
                f" timestep schedules. Please check whether you are using the correct scheduler."
            )
        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
        timesteps = scheduler.timesteps
        num_inference_steps = len(timesteps)
    elif sigmas is not None:
        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
        if not accept_sigmas:
            raise ValueError(
                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
                f" sigmas schedules. Please check whether you are using the correct scheduler."
            )
        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
        timesteps = scheduler.timesteps
        num_inference_steps = len(timesteps)
    else:
        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
        timesteps = scheduler.timesteps
    return timesteps, num_inference_steps


class FluxPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFileMixin):
    def __init__(
            self,
            scheduler: FlowMatchEulerDiscreteScheduler,
            vae: AutoencoderKL,
            text_encoder: CLIPTextModel,
            tokenizer: CLIPTokenizer,
            text_encoder_2: T5EncoderModel,
            tokenizer_2: T5TokenizerFast,
            transformer: FluxTransformer2DModel,
    ):
        super().__init__()

        self.register_modules(
            vae=vae,
            text_encoder=text_encoder,
            text_encoder_2=text_encoder_2,
            tokenizer=tokenizer,
            tokenizer_2=tokenizer_2,
            transformer=transformer,
            scheduler=scheduler,
        )
        self.vae_scale_factor = (
            2 ** (len(self.vae.config.block_out_channels)) if hasattr(self, "vae") and self.vae is not None else 16
        )
        self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
        self.tokenizer_max_length = (
            self.tokenizer.model_max_length if hasattr(self, "tokenizer") and self.tokenizer is not None else 77
        )
        self.default_sample_size = 64

    def _get_t5_prompt_embeds(
            self,
            prompt: Union[str, List[str]] = None,
            num_images_per_prompt: int = 1,
            max_sequence_length: int = 512,
            device: Optional[torch.device] = None,
            dtype: Optional[torch.dtype] = None,
    ):
        device = device or self._execution_device
        dtype = dtype or self.text_encoder.dtype

        prompt = [prompt] if isinstance(prompt, str) else prompt
        batch_size = len(prompt)

        text_inputs = self.tokenizer_2(
            prompt,
            padding="max_length",
            max_length=max_sequence_length,
            truncation=True,
            return_length=False,
            return_overflowing_tokens=False,
            return_tensors="pt",
        )
        text_input_ids = text_inputs.input_ids
        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").input_ids

        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1: -1])
            logger.warning(
                "The following part of your input was truncated because `max_sequence_length` is set to "
                f" {max_sequence_length} tokens: {removed_text}"
            )

        prompt_embeds = self.text_encoder_2(text_input_ids.to(device), output_hidden_states=False)[0]

        dtype = self.text_encoder_2.dtype
        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)

        _, seq_len, _ = prompt_embeds.shape

        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)

        return prompt_embeds

    def _get_clip_prompt_embeds(
            self,
            prompt: Union[str, List[str]],
            num_images_per_prompt: int = 1,
            device: Optional[torch.device] = None,
    ):
        device = device or self._execution_device

        prompt = [prompt] if isinstance(prompt, str) else prompt
        batch_size = len(prompt)

        text_inputs = self.tokenizer(
            prompt,
            padding="max_length",
            max_length=self.tokenizer_max_length,
            truncation=True,
            return_overflowing_tokens=False,
            return_length=False,
            return_tensors="pt",
        )

        text_input_ids = text_inputs.input_ids
        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1: -1])
            logger.warning(
                "The following part of your input was truncated because CLIP can only handle sequences up to"
                f" {self.tokenizer_max_length} tokens: {removed_text}"
            )
        prompt_embeds = self.text_encoder(text_input_ids.to(device), output_hidden_states=False)

        # Use pooled output of CLIPTextModel
        prompt_embeds = prompt_embeds.pooler_output
        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)

        # duplicate text embeddings for each generation per prompt, using mps friendly method
        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt)
        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)

        return prompt_embeds

    def encode_prompt(
            self,
            prompt: Union[str, List[str]],
            prompt_2: Union[str, List[str]],
            device: Optional[torch.device] = None,
            num_images_per_prompt: int = 1,
            prompt_embeds: Optional[torch.FloatTensor] = None,
            pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
            max_sequence_length: int = 512,
            lora_scale: Optional[float] = None,
    ):
        device = device or self._execution_device

        # set lora scale so that monkey patched LoRA
        # function of text encoder can correctly access it
        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
            self._lora_scale = lora_scale

            # dynamically adjust the LoRA scale
            if self.text_encoder is not None and USE_PEFT_BACKEND:
                scale_lora_layers(self.text_encoder, lora_scale)
            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
                scale_lora_layers(self.text_encoder_2, lora_scale)

        prompt = [prompt] if isinstance(prompt, str) else prompt

        if prompt_embeds is None:
            prompt_2 = prompt_2 or prompt
            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2

            # We only use the pooled prompt output from the CLIPTextModel
            pooled_prompt_embeds = self._get_clip_prompt_embeds(
                prompt=prompt,
                device=device,
                num_images_per_prompt=num_images_per_prompt,
            )
            prompt_embeds = self._get_t5_prompt_embeds(
                prompt=prompt_2,
                num_images_per_prompt=num_images_per_prompt,
                max_sequence_length=max_sequence_length,
                device=device,
            )

        if self.text_encoder is not None:
            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
                # Retrieve the original scale by scaling back the LoRA layers
                unscale_lora_layers(self.text_encoder, lora_scale)

        if self.text_encoder_2 is not None:
            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
                # Retrieve the original scale by scaling back the LoRA layers
                unscale_lora_layers(self.text_encoder_2, lora_scale)

        dtype = self.text_encoder.dtype if self.text_encoder is not None else self.transformer.dtype
        text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)

        return prompt_embeds, pooled_prompt_embeds, text_ids

    # Copied from diffusers.pipelines.stable_diffusion_3.pipeline_stable_diffusion_3_inpaint.StableDiffusion3InpaintPipeline._encode_vae_image
    def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
        if isinstance(generator, list):
            image_latents = [
                retrieve_latents(self.vae.encode(image[i: i + 1]), generator=generator[i])
                for i in range(image.shape[0])
            ]
            image_latents = torch.cat(image_latents, dim=0)
        else:
            image_latents = retrieve_latents(self.vae.encode(image), generator=generator)

        image_latents = (image_latents - self.vae.config.shift_factor) * self.vae.config.scaling_factor

        return image_latents

    def check_inputs(
            self,
            prompt,
            prompt_2,
            height,
            width,
            prompt_embeds=None,
            pooled_prompt_embeds=None,
            callback_on_step_end_tensor_inputs=None,
            max_sequence_length=None,
    ):
        if height % 8 != 0 or width % 8 != 0:
            raise ValueError(f"`height` and `width` have to be divisible by 8 but are {height} and {width}.")

        if prompt is not None and prompt_embeds is not None:
            raise ValueError(
                f"Cannot forward both `prompt`: {prompt} and `prompt_embeds`: {prompt_embeds}. Please make sure to"
                " only forward one of the two."
            )
        elif prompt_2 is not None and prompt_embeds is not None:
            raise ValueError(
                f"Cannot forward both `prompt_2`: {prompt_2} and `prompt_embeds`: {prompt_embeds}. Please make sure to"
                " only forward one of the two."
            )
        elif prompt is None and prompt_embeds is None:
            raise ValueError(
                "Provide either `prompt` or `prompt_embeds`. Cannot leave both `prompt` and `prompt_embeds` undefined."
            )
        elif prompt is not None and (not isinstance(prompt, str) and not isinstance(prompt, list)):
            raise ValueError(f"`prompt` has to be of type `str` or `list` but is {type(prompt)}")
        elif prompt_2 is not None and (not isinstance(prompt_2, str) and not isinstance(prompt_2, list)):
            raise ValueError(f"`prompt_2` has to be of type `str` or `list` but is {type(prompt_2)}")

        if prompt_embeds is not None and pooled_prompt_embeds is None:
            raise ValueError(
                "If `prompt_embeds` are provided, `pooled_prompt_embeds` also have to be passed. Make sure to generate `pooled_prompt_embeds` from the same text encoder that was used to generate `prompt_embeds`."
            )

        if max_sequence_length is not None and max_sequence_length > 512:
            raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")

    @staticmethod
    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
        latent_image_ids = torch.zeros(height // 2, width // 2, 3)
        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height // 2)[:, None]
        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width // 2)[None, :]
        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
        latent_image_ids = latent_image_ids.reshape(
            latent_image_id_height * latent_image_id_width, latent_image_id_channels
        )
        return latent_image_ids.to(device=device, dtype=dtype)

    @staticmethod
    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
        latents = latents.permute(0, 2, 4, 1, 3, 5)
        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
        return latents

    @staticmethod
    def _unpack_latents(latents, height, width, vae_scale_factor):
        batch_size, num_patches, channels = latents.shape

        height = height // vae_scale_factor
        width = width // vae_scale_factor

        latents = latents.view(batch_size, height, width, channels // 4, 2, 2)
        latents = latents.permute(0, 3, 1, 4, 2, 5)

        latents = latents.reshape(batch_size, channels // (2 * 2), height * 2, width * 2)

        return latents

    def enable_vae_slicing(self):
        r"""
        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.
        """
        self.vae.enable_slicing()

    def disable_vae_slicing(self):
        r"""
        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
        computing decoding in one step.
        """
        self.vae.disable_slicing()

    def enable_vae_tiling(self):
        r"""
        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.
        """
        self.vae.enable_tiling()

    def disable_vae_tiling(self):
        r"""
        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
        computing decoding in one step.
        """
        self.vae.disable_tiling()

    def prepare_latents(
            self,
            batch_size,
            num_channels_latents,
            height,
            width,
            dtype,
            device,
            generator,
            subject_image,
            condition_image,
            latents=None,
            cond_number=1,
            sub_number=1
    ):
        height_cond = 2 * (self.cond_size // self.vae_scale_factor)
        width_cond = 2 * (self.cond_size // self.vae_scale_factor)
        height = 2 * (int(height) // self.vae_scale_factor)  
        width = 2 * (int(width) // self.vae_scale_factor)

        shape = (batch_size, num_channels_latents, height, width)  # 1 16 106 80
        noise_latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)  
        noise_latents = self._pack_latents(noise_latents, batch_size, num_channels_latents, height, width)
        noise_latent_image_ids, cond_latent_image_ids = resize_position_encoding(
                batch_size,
                height,
                width,
                height_cond,
                width_cond,
                device,
                dtype,
            )
        
        latents_to_concat = []
        latents_ids_to_concat = [noise_latent_image_ids]
        
        # subject
        if subject_image is not None:
            shape_subject = (batch_size, num_channels_latents, height_cond*sub_number, width_cond)  
            subject_image = subject_image.to(device=device, dtype=dtype)
            subject_image_latents = self._encode_vae_image(image=subject_image, generator=generator)
            subject_latents = self._pack_latents(subject_image_latents, batch_size, num_channels_latents, height_cond*sub_number, width_cond)
            mask2 = torch.zeros(shape_subject, device=device, dtype=dtype)
            mask2 = self._pack_latents(mask2, batch_size, num_channels_latents, height_cond*sub_number, width_cond)
            latent_subject_ids = prepare_latent_subject_ids(height_cond, width_cond, device, dtype)
            latent_subject_ids[:, 1] += 64  # fixed offset
            subject_latent_image_ids = torch.concat([latent_subject_ids for _ in range(sub_number)], dim=-2)
            latents_to_concat.append(subject_latents)
            latents_ids_to_concat.append(subject_latent_image_ids)
            
        # spatial
        if condition_image is not None:
            shape_cond = (batch_size, num_channels_latents, height_cond*cond_number, width_cond)  
            condition_image = condition_image.to(device=device, dtype=dtype)
            image_latents = self._encode_vae_image(image=condition_image, generator=generator)
            cond_latents = self._pack_latents(image_latents, batch_size, num_channels_latents, height_cond*cond_number, width_cond)
            mask3 = torch.zeros(shape_cond, device=device, dtype=dtype)
            mask3 = self._pack_latents(mask3, batch_size, num_channels_latents, height_cond*cond_number, width_cond) 
            cond_latent_image_ids = cond_latent_image_ids
            cond_latent_image_ids = torch.concat([cond_latent_image_ids for _ in range(cond_number)], dim=-2)
            latents_ids_to_concat.append(cond_latent_image_ids)
            latents_to_concat.append(cond_latents)

        cond_latents = torch.concat(latents_to_concat, dim=-2)
        latent_image_ids = torch.concat(latents_ids_to_concat, dim=-2)
        return cond_latents, latent_image_ids, noise_latents

    @property
    def guidance_scale(self):
        return self._guidance_scale

    @property
    def joint_attention_kwargs(self):
        return self._joint_attention_kwargs

    @property
    def num_timesteps(self):
        return self._num_timesteps

    @property
    def interrupt(self):
        return self._interrupt

    @torch.no_grad()
    def __call__(
            self,
            prompt: Union[str, List[str]] = None,
            prompt_2: Optional[Union[str, List[str]]] = None,
            height: Optional[int] = None,
            width: Optional[int] = None,
            num_inference_steps: int = 28,
            timesteps: List[int] = None,
            guidance_scale: float = 3.5,
            num_images_per_prompt: Optional[int] = 1,
            generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
            latents: Optional[torch.FloatTensor] = None,
            prompt_embeds: Optional[torch.FloatTensor] = None,
            pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
            output_type: Optional[str] = "pil",
            return_dict: bool = True,
            joint_attention_kwargs: Optional[Dict[str, Any]] = None,
            callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = None,
            callback_on_step_end_tensor_inputs: List[str] = ["latents"],
            max_sequence_length: int = 512,
            spatial_images=[],
            subject_images=[],
            cond_size=512,
    ):

        height = height or self.default_sample_size * self.vae_scale_factor
        width = width or self.default_sample_size * self.vae_scale_factor
        self.cond_size = cond_size
        
        # 1. Check inputs. Raise error if not correct
        self.check_inputs(
            prompt,
            prompt_2,
            height,
            width,
            prompt_embeds=prompt_embeds,
            pooled_prompt_embeds=pooled_prompt_embeds,
            callback_on_step_end_tensor_inputs=callback_on_step_end_tensor_inputs,
            max_sequence_length=max_sequence_length,
        )

        self._guidance_scale = guidance_scale
        self._joint_attention_kwargs = joint_attention_kwargs
        self._interrupt = False
        
        cond_number = len(spatial_images)
        sub_number = len(subject_images)
        
        if sub_number > 0:
            subject_image_ls = []
            for subject_image in subject_images:
                w, h = subject_image.size[:2]
                scale = self.cond_size / max(h, w)
                new_h, new_w = int(h * scale), int(w * scale)
                subject_image = self.image_processor.preprocess(subject_image, height=new_h, width=new_w)
                subject_image = subject_image.to(dtype=torch.float32)
                pad_h = cond_size - subject_image.shape[-2]
                pad_w = cond_size - subject_image.shape[-1]
                subject_image = pad(
                    subject_image,
                    padding=(int(pad_w / 2), int(pad_h / 2), int(pad_w / 2), int(pad_h / 2)),
                    fill=0
                )
                subject_image_ls.append(subject_image)
            subject_image = torch.concat(subject_image_ls, dim=-2)
        else:
            subject_image = None
        
        if cond_number > 0:
            condition_image_ls = []
            for img in spatial_images:
                print(img)
                condition_image = self.image_processor.preprocess(img, height=self.cond_size, width=self.cond_size)
                condition_image = condition_image.to(dtype=torch.float32)
                condition_image_ls.append(condition_image)
            condition_image = torch.concat(condition_image_ls, dim=-2)
        else:
            condition_image = None
        
        # 2. Define call parameters
        if prompt is not None and isinstance(prompt, str):
            batch_size = 1
        elif prompt is not None and isinstance(prompt, list):
            batch_size = len(prompt)
        else:
            batch_size = prompt_embeds.shape[0]

        device = self._execution_device

        lora_scale = (
            self.joint_attention_kwargs.get("scale", None) if self.joint_attention_kwargs is not None else None
        )
        (
            prompt_embeds,
            pooled_prompt_embeds,
            text_ids,
        ) = self.encode_prompt(
            prompt=prompt,
            prompt_2=prompt_2,
            prompt_embeds=prompt_embeds,
            pooled_prompt_embeds=pooled_prompt_embeds,
            device=device,
            num_images_per_prompt=num_images_per_prompt,
            max_sequence_length=max_sequence_length,
            lora_scale=lora_scale,
        )

        # 4. Prepare latent variables
        num_channels_latents = self.transformer.config.in_channels // 4  # 16
        cond_latents, latent_image_ids, noise_latents = self.prepare_latents(
            batch_size * num_images_per_prompt,
            num_channels_latents,
            height,
            width,
            prompt_embeds.dtype,
            device,
            generator,
            subject_image,
            condition_image,
            latents,
            cond_number,
            sub_number
        )
        latents = noise_latents
        # 5. Prepare timesteps
        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps)
        image_seq_len = latents.shape[1]
        mu = calculate_shift(
            image_seq_len,
            self.scheduler.config.base_image_seq_len,
            self.scheduler.config.max_image_seq_len,
            self.scheduler.config.base_shift,
            self.scheduler.config.max_shift,
        )
        timesteps, num_inference_steps = retrieve_timesteps(
            self.scheduler,
            num_inference_steps,
            device,
            timesteps,
            sigmas,
            mu=mu,
        )
        num_warmup_steps = max(len(timesteps) - num_inference_steps * self.scheduler.order, 0)
        self._num_timesteps = len(timesteps)

        # handle guidance
        if self.transformer.config.guidance_embeds:
            guidance = torch.full([1], guidance_scale, device=device, dtype=torch.float32)
            guidance = guidance.expand(latents.shape[0])
        else:
            guidance = None

        ## Caching conditions
        # clean the cache
        try:
            for name, attn_processor in self.transformer.attn_processors.items():
                attn_processor.bank_kv.clear()
        except:
            pass
        # cache with warmup latents
        t = torch.tensor([timesteps[0]], device=device)
        timestep = t.expand(cond_latents.shape[0]).to(latents.dtype)
        warmup_image_ids = latent_image_ids[latents.shape[1]:, :]
        _ = self.transformer(
                    hidden_states=cond_latents,  
                    timestep=torch.ones_like(timestep) * 0,
                    guidance=guidance,
                    pooled_projections=pooled_prompt_embeds,
                    encoder_hidden_states=prompt_embeds,
                    txt_ids=text_ids,
                    img_ids=warmup_image_ids,
                    joint_attention_kwargs=self.joint_attention_kwargs,
                    return_dict=False,
                )[0]
        
        del cond_latents, spatial_images, condition_image, condition_image_ls, img, _
        torch.cuda.empty_cache()

        # 6. Denoising loop
        with self.progress_bar(total=num_inference_steps) as progress_bar:
            for i, t in enumerate(timesteps):
                if self.interrupt:
                    continue

                # broadcast to batch dimension in a way that's compatible with ONNX/Core ML
                timestep = t.expand(latents.shape[0]).to(latents.dtype)
                noise_pred = self.transformer(
                    hidden_states=latents,
                    timestep=timestep / 1000,
                    guidance=guidance,
                    pooled_projections=pooled_prompt_embeds,
                    encoder_hidden_states=prompt_embeds,
                    txt_ids=text_ids,
                    img_ids=latent_image_ids,
                    joint_attention_kwargs=self.joint_attention_kwargs,
                    return_dict=False,
                )[0]

                # compute the previous noisy sample x_t -> x_t-1
                latents_dtype = latents.dtype
                latents = self.scheduler.step(noise_pred, t, latents, return_dict=False)[0]

                if latents.dtype != latents_dtype:
                    if torch.backends.mps.is_available():
                        # some platforms (eg. apple mps) misbehave due to a pytorch bug: https://github.com/pytorch/pytorch/pull/99272
                        latents = latents.to(latents_dtype)

                if callback_on_step_end is not None:
                    callback_kwargs = {}
                    for k in callback_on_step_end_tensor_inputs:
                        callback_kwargs[k] = locals()[k]
                    callback_outputs = callback_on_step_end(self, i, t, callback_kwargs)

                    latents = callback_outputs.pop("latents", latents)
                    prompt_embeds = callback_outputs.pop("prompt_embeds", prompt_embeds)

                # call the callback, if provided
                if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
                    progress_bar.update()

                if XLA_AVAILABLE:
                    xm.mark_step()

        if output_type == "latent":
            image = latents
        else:
            latents = self._unpack_latents(latents, height, width, self.vae_scale_factor)
            latents = (latents / self.vae.config.scaling_factor) + self.vae.config.shift_factor
            image = self.vae.decode(latents.to(dtype=self.vae.dtype), return_dict=False)[0]
            image = self.image_processor.postprocess(image, output_type=output_type)

        # Offload all models
        self.maybe_free_model_hooks()

        if not return_dict:
            return (image,)

        return FluxPipelineOutput(images=image)