snt_pipeline.py

# Copyright 2023 The InstructPix2Pix Authors and The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import inspect
import warnings
from typing import Callable, List, Optional, Union

import PIL
import torch
from transformers import CLIPImageProcessor

from diffusers.image_processor import VaeImageProcessor
from diffusers.loaders import LoraLoaderMixin, TextualInversionLoaderMixin
from diffusers.utils import (
    deprecate,
    is_accelerate_available,
    is_accelerate_version,
    logging,
)

try: 
    from diffusers.utils import randn_tensor
except ImportError:
    from diffusers.utils.torch_utils import randn_tensor


from diffusers.pipelines.pipeline_utils import DiffusionPipeline
from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker

from .sd_model import SDModel


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

from typing import Callable, List, Optional, Union
import PIL

from transformers import CLIPImageProcessor

from diffusers.image_processor import VaeImageProcessor









# from hydra.utils import instantiate

from einops import rearrange, repeat


class ShowNotTellPipeline(DiffusionPipeline, TextualInversionLoaderMixin, LoraLoaderMixin):
    r"""
    Pipeline for pixel-level image editing by following text instructions. Based on Stable Diffusion.

    This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods the
    library implements for all the pipelines (such as downloading or saving, running on a particular device, etc.)

    In addition the pipeline inherits the following loading methods:
        - *Textual-Inversion*: [`loaders.TextualInversionLoaderMixin.load_textual_inversion`]
        - *LoRA*: [`loaders.LoraLoaderMixin.load_lora_weights`]

    as well as the following saving methods:
        - *LoRA*: [`loaders.LoraLoaderMixin.save_lora_weights`]

    Args:
        vae ([`AutoencoderKL`]):
            Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations.
        text_encoder ([`CLIPTextModel`]):
            Frozen text-encoder. Stable Diffusion uses the text portion of
            [CLIP](https://huggingface.co/docs/transformers/model_doc/clip#transformers.CLIPTextModel), specifically
            the [clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14) variant.
        tokenizer (`CLIPTokenizer`):
            Tokenizer of class
            [CLIPTokenizer](https://huggingface.co/docs/transformers/v4.21.0/en/model_doc/clip#transformers.CLIPTokenizer).
        unet ([`UNet2DConditionModel`]): Conditional U-Net architecture to denoise the encoded image latents.
        scheduler ([`SchedulerMixin`]):
            A scheduler to be used in combination with `unet` to denoise the encoded image latents. Can be one of
            [`DDIMScheduler`], [`LMSDiscreteScheduler`], or [`PNDMScheduler`].
        safety_checker ([`StableDiffusionSafetyChecker`]):
            Classification module that estimates whether generated images could be considered offensive or harmful.
            Please, refer to the [model card](https://huggingface.co/runwayml/stable-diffusion-v1-5) for details.
        feature_extractor ([`CLIPImageProcessor`]):
            Model that extracts features from generated images to be used as inputs for the `safety_checker`.
    """
    _optional_components = ["safety_checker", "feature_extractor"]

    def __init__(
        self,
        # cfg: SDModelConfig,
        model: SDModel,
        safety_checker: StableDiffusionSafetyChecker = None, 
        feature_extractor: CLIPImageProcessor = None, 
        requires_safety_checker: bool = False,        
    ):
        super().__init__()
        # self.model.cfg = cfg
        self.register_modules(model=model, safety_checker=safety_checker, feature_extractor=feature_extractor)
        # self.register_to_config(cfg=dataclasses.asdict(cfg))
        
        self.model.vae_scale_factor = 2 ** (len(self.model.vae.config.block_out_channels) - 1)
        self.image_processor = VaeImageProcessor(vae_scale_factor=self.model.vae_scale_factor)
        self.register_to_config(requires_safety_checker=requires_safety_checker)

    @torch.no_grad()
    def __call__(        
        self,
        prompts,
        image,
        num_inference_steps: int = 100,
        guidance_scale: float = 7.5,
        image_guidance_scale: float = 1.5,
        negative_prompt: Optional[Union[str, List[str]]] = None,
        num_images_per_prompt: Optional[int] = 1,
        eta: float = 0.0,
        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
        latents: Optional[torch.FloatTensor] = None,
        prompt_embeds: Optional[torch.FloatTensor] = None,
        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
        output_type: Optional[str] = "pil",
        return_dict: bool = True,
        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
        callback_steps: int = 1,):
        
        if isinstance(prompts, str):
            prompts = [prompts]
        if isinstance(prompts, list):
            input_ids = self.fancy_get_input_ids(prompts, self.model.text_encoder.device) # TODO see if reshaping needed to match train dataloader
        else:
            input_ids = prompts
        
        if isinstance(image, PIL.Image.Image):
            image = [image]
        if isinstance(image, list):
            preprocessed_images = self.image_processor.preprocess(image)
        else:
            preprocessed_images = image
            
        batch_size = input_ids.shape[0]
        
        # device = self._execution_device
        device = self.model.text_encoder.device # TODO figure out execution device stuff
        # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
        # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`
        # corresponds to doing no classifier free guidance.
        do_classifier_free_guidance = guidance_scale > 1.0 and image_guidance_scale >= 1.0
        # check if scheduler is in sigmas space
        scheduler_is_in_sigma_space = hasattr(self.model.noise_scheduler, "sigmas")
        
        
        prompt_embeds = self.encode_prompt_batch(input_ids, batch_size, device, num_images_per_prompt, do_classifier_free_guidance, negative_prompt, prompt_embeds, negative_prompt_embeds)
        
        # 4. set timesteps
        self.model.noise_scheduler.set_timesteps(num_inference_steps, device=device)
        timesteps = self.model.noise_scheduler.timesteps
        
        # 5. Prepare Image latents
        image_latents = self.prepare_image_latents(
            preprocessed_images,
            batch_size,
            num_images_per_prompt,
            prompt_embeds.dtype,
            device,
            do_classifier_free_guidance,
            generator,
        )
        
        height, width = image_latents.shape[-2:]
        height = height * self.model.vae_scale_factor
        width = width * self.model.vae_scale_factor
        
        # 6. Prepare latent variables
        num_channels_latents = self.model.vae.config.latent_channels
        
        latents = self.prepare_latents(
            batch_size * num_images_per_prompt,
            num_channels_latents,
            height,
            width,
            prompt_embeds.dtype,
            device,
            generator,
            latents,
        )
        
        # 7. Check that shapes of latents and image match the UNet channels
        num_channels_image = image_latents.shape[1]
        if num_channels_latents + num_channels_image != self.model.unet.config.in_channels:
            raise ValueError(
                f"Incorrect configuration settings! The config of `pipeline.model.unet`: {self.model.unet.config} expects"
                f" {self.model.unet.config.in_channels} but received `num_channels_latents`: {num_channels_latents} +"
                f" `num_channels_image`: {num_channels_image} "
                f" = {num_channels_latents+num_channels_image}. Please verify the config of"
                " `pipeline.model.unet` or your `image` input."
            )

        # 8. Prepare extra step kwargs. TODO: Logic should ideally just be moved out of the pipeline
        extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)

        # 9. Denoising loop
        num_warmup_steps = len(timesteps) - num_inference_steps * self.model.noise_scheduler.order
        with self.progress_bar(total=num_inference_steps) as progress_bar:
            for i, t in enumerate(timesteps):

                
                
                # Expand the latents if we are doing classifier free guidance.
                # The latents are expanded 3 times because for pix2pix the guidance\
                # is applied for both the text and the input image.
                latent_model_input = torch.cat([latents] * 3) if do_classifier_free_guidance else latents
                # if i == 0:    
                #     if self.model.cfg.image_positional_encoding_type is not None: 
                #         third = latents.shape[0]//3 
                #         cond_latents = latents[third:2*third]
                #         cond_latents = rearrange(cond_latents, 'b c (s h) w -> (b s) c h w', s=self.model.cfg.sequence_length)
                #         cond_latents = self.model.apply_image_positional_encoding(cond_latents, self.model.cfg.sequence_length)
                #         cond_latents = rearrange(cond_latents, '(b s) c h w -> b c (s h) w', s=self.model.cfg.sequence_length)
                #         latents[third:2*third] = cond_latents
                
                # concat latents, image_latents in the channel dimension
                scaled_latent_model_input = self.model.noise_scheduler.scale_model_input(latent_model_input, t)

                scaled_latent_model_input = torch.cat([scaled_latent_model_input, image_latents], dim=1)

                # predict the noise residual
                noise_pred = self.model.unet(
                    scaled_latent_model_input, t, encoder_hidden_states=prompt_embeds, return_dict=False
                )[0]

                # Hack:
                # For karras style schedulers the model does classifer free guidance using the
                # predicted_original_sample instead of the noise_pred. So we need to compute the
                # predicted_original_sample here if we are using a karras style scheduler.
                if scheduler_is_in_sigma_space:
                    step_index = (self.model.noise_scheduler.timesteps == t).nonzero().item()
                    sigma = self.model.noise_scheduler.sigmas[step_index]
                    noise_pred = latent_model_input - sigma * noise_pred

                # perform guidance
                if do_classifier_free_guidance:
                    noise_pred_text, noise_pred_image, noise_pred_uncond = noise_pred.chunk(3)
                    noise_pred = (
                        noise_pred_uncond
                        + guidance_scale * (noise_pred_text - noise_pred_image)
                        + image_guidance_scale * (noise_pred_image - noise_pred_uncond)
                    )

                # Hack:
                # For karras style schedulers the model does classifer free guidance using the
                # predicted_original_sample instead of the noise_pred. But the scheduler.step function
                # expects the noise_pred and computes the predicted_original_sample internally. So we
                # need to overwrite the noise_pred here such that the value of the computed
                # predicted_original_sample is correct.
                if scheduler_is_in_sigma_space:
                    noise_pred = (noise_pred - latents) / (-sigma)

                # compute the previous noisy sample x_t -> x_t-1
                latents = self.model.noise_scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]
                


                # call the callback, if provided
                if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.model.noise_scheduler.order == 0):
                    progress_bar.update()
                    if callback is not None and i % callback_steps == 0:
                        callback(i, t, latents)

        if not output_type == "latent":
            latents = rearrange(latents, 'b c (s h) w -> (b s) c h w', s=self.model.cfg.sequence_length) # these are image latents, so sequence_length instead of text_sequence_length
            image = self.model.vae.decode(latents / self.model.vae.config.scaling_factor, return_dict=False)[0]
            # image, has_nsfw_concept = self.run_safety_checker(image, device, prompt_embeds.dtype)
        else:
            image = latents

        has_nsfw_concept = None
        do_denormalize = [True] * image.shape[0]
        # if has_nsfw_concept is None:
        #     do_denormalize = [True] * image.shape[0]
        # else:
        #     do_denormalize = [not has_nsfw for has_nsfw in has_nsfw_concept]

        image = self.image_processor.postprocess(image, output_type=output_type, do_denormalize=do_denormalize)

        # Offload last model to CPU
        if hasattr(self, "final_offload_hook") and self.final_offload_hook is not None:
            self.final_offload_hook.offload()

        if not return_dict:
            return (image, has_nsfw_concept)

        return StableDiffusionPipelineOutput(images=image, nsfw_content_detected=has_nsfw_concept)
    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_sequential_cpu_offload
    def enable_sequential_cpu_offload(self, gpu_id=0):
        r"""
        Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet,
        text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a
        `torch.device('meta') and loaded to GPU only when their specific submodule has its `forward` method called.
        Note that offloading happens on a submodule basis. Memory savings are higher than with
        `enable_model_cpu_offload`, but performance is lower.
        """
        if is_accelerate_available() and is_accelerate_version(">=", "0.14.0"):
            from accelerate import cpu_offload
        else:
            raise ImportError("`enable_sequential_cpu_offload` requires `accelerate v0.14.0` or higher")

        device = torch.device(f"cuda:{gpu_id}")

        if self.device.type != "cpu":
            self.to("cpu", silence_dtype_warnings=True)
            torch.cuda.empty_cache()  # otherwise we don't see the memory savings (but they probably exist)

        for cpu_offloaded_model in [self.model.unet, self.model.text_encoder, self.model.vae]:
            cpu_offload(cpu_offloaded_model, device)

        if self.safety_checker is not None:
            cpu_offload(self.safety_checker, execution_device=device, offload_buffers=True)

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.enable_model_cpu_offload
    def enable_model_cpu_offload(self, gpu_id=0):
        r"""
        Offloads all models to CPU using accelerate, reducing memory usage with a low impact on performance. Compared
        to `enable_sequential_cpu_offload`, this method moves one whole model at a time to the GPU when its `forward`
        method is called, and the model remains in GPU until the next model runs. Memory savings are lower than with
        `enable_sequential_cpu_offload`, but performance is much better due to the iterative execution of the `unet`.
        """
        if is_accelerate_available() and is_accelerate_version(">=", "0.17.0.dev0"):
            from accelerate import cpu_offload_with_hook
        else:
            raise ImportError("`enable_model_cpu_offload` requires `accelerate v0.17.0` or higher.")

        device = torch.device(f"cuda:{gpu_id}")

        if self.device.type != "cpu":
            self.to("cpu", silence_dtype_warnings=True)
            torch.cuda.empty_cache()  # otherwise we don't see the memory savings (but they probably exist)

        hook = None
        for cpu_offloaded_model in [self.model.text_encoder, self.model.unet, self.model.vae]:
            _, hook = cpu_offload_with_hook(cpu_offloaded_model, device, prev_module_hook=hook)

        if self.safety_checker is not None:
            _, hook = cpu_offload_with_hook(self.safety_checker, device, prev_module_hook=hook)

        # We'll offload the last model manually.
        self.final_offload_hook = hook

    @property
    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline._execution_device
    def _execution_device(self):
        r"""
        Returns the device on which the pipeline's models will be executed. After calling
        `pipeline.enable_sequential_cpu_offload()` the execution device can only be inferred from Accelerate's module
        hooks.
        """
        if not hasattr(self.model.unet, "_hf_hook"):
            return self.device
        for module in self.model.unet.modules():
            if (
                hasattr(module, "_hf_hook")
                and hasattr(module._hf_hook, "execution_device")
                and module._hf_hook.execution_device is not None
            ):
                return torch.device(module._hf_hook.execution_device)
        return self.device

    def _encode_prompt(
        self,
        prompt,
        device,
        num_images_per_prompt,
        do_classifier_free_guidance,
        negative_prompt=None,
        prompt_embeds: Optional[torch.FloatTensor] = None,
        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
    ):
        r"""
        Encodes the prompt into text encoder hidden states.

        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
            do_classifier_free_guidance (`bool`):
                whether to use classifier free guidance or not
            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`).
            prompt_embeds (`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 (`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.
        """
        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]

        if prompt_embeds is None:
            # textual inversion: procecss multi-vector tokens if necessary
            if isinstance(self, TextualInversionLoaderMixin):
                prompt = self.maybe_convert_prompt(prompt, self.model.tokenizer)

            text_inputs = self.model.tokenizer(
                prompt,
                padding="max_length",
                max_length=self.model.tokenizer.model_max_length,
                truncation=True,
                return_tensors="pt",
            )
            text_input_ids = text_inputs.input_ids
            untruncated_ids = self.model.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.model.tokenizer.batch_decode(
                    untruncated_ids[:, self.model.tokenizer.model_max_length - 1 : -1]
                )
                logger.warning(
                    "The following part of your input was truncated because CLIP can only handle sequences up to"
                    f" {self.model.tokenizer.model_max_length} tokens: {removed_text}"
                )

            if hasattr(self.model.text_encoder.config, "use_attention_mask") and self.model.text_encoder.config.use_attention_mask:
                attention_mask = text_inputs.attention_mask.to(device)
            else:
                attention_mask = None

            prompt_embeds = self.model.text_encoder(
                text_input_ids.to(device),
                attention_mask=attention_mask,
            )
            prompt_embeds = prompt_embeds[0]

        prompt_embeds = prompt_embeds.to(dtype=self.model.text_encoder.dtype, device=device)

        bs_embed, seq_len, _ = prompt_embeds.shape
        # duplicate text embeddings 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(bs_embed * num_images_per_prompt, seq_len, -1)

        # get unconditional embeddings for classifier free guidance
        if do_classifier_free_guidance and negative_prompt_embeds is None:
            uncond_tokens: List[str]
            if negative_prompt is None:
                uncond_tokens = [""] * batch_size
            elif type(prompt) is not type(negative_prompt):
                raise TypeError(
                    f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
                    f" {type(prompt)}."
                )
            elif isinstance(negative_prompt, str):
                uncond_tokens = [negative_prompt]
            elif batch_size != len(negative_prompt):
                raise ValueError(
                    f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
                    f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
                    " the batch size of `prompt`."
                )
            else:
                uncond_tokens = negative_prompt

            # textual inversion: procecss multi-vector tokens if necessary
            if isinstance(self, TextualInversionLoaderMixin):
                uncond_tokens = self.maybe_convert_prompt(uncond_tokens, self.model.tokenizer)

            max_length = prompt_embeds.shape[1]
            uncond_input = self.model.tokenizer(
                uncond_tokens,
                padding="max_length",
                max_length=max_length,
                truncation=True,
                return_tensors="pt",
            )

            if hasattr(self.model.text_encoder.config, "use_attention_mask") and self.model.text_encoder.config.use_attention_mask:
                attention_mask = uncond_input.attention_mask.to(device)
            else:
                attention_mask = None

            negative_prompt_embeds = self.model.text_encoder(
                uncond_input.input_ids.to(device),
                attention_mask=attention_mask,
            )
            negative_prompt_embeds = negative_prompt_embeds[0]

        if do_classifier_free_guidance:
            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
            seq_len = negative_prompt_embeds.shape[1]

            negative_prompt_embeds = negative_prompt_embeds.to(dtype=self.model.text_encoder.dtype, device=device)

            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)

            # For classifier free guidance, we need to do two forward passes.
            # Here we concatenate the unconditional and text embeddings into a single batch
            # to avoid doing two forward passes
            # pix2pix has two  negative embeddings, and unlike in other pipelines latents are ordered [prompt_embeds, negative_prompt_embeds, negative_prompt_embeds]
            prompt_embeds = torch.cat([prompt_embeds, negative_prompt_embeds, negative_prompt_embeds])

        return prompt_embeds

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.run_safety_checker
    def run_safety_checker(self, image, device, dtype):
        if self.safety_checker is None:
            has_nsfw_concept = None
        else:
            if torch.is_tensor(image):
                feature_extractor_input = self.image_processor.postprocess(image, output_type="pil")
            else:
                feature_extractor_input = self.image_processor.numpy_to_pil(image)
            safety_checker_input = self.feature_extractor(feature_extractor_input, return_tensors="pt").to(device)
            image, has_nsfw_concept = self.safety_checker(
                images=image, clip_input=safety_checker_input.pixel_values.to(dtype)
            )
        return image, has_nsfw_concept

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
    def prepare_extra_step_kwargs(self, generator, eta):
        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
        # eta corresponds to η in DDIM paper: https://arxiv.org/abs/2010.02502
        # and should be between [0, 1]

        accepts_eta = "eta" in set(inspect.signature(self.model.noise_scheduler.step).parameters.keys())
        extra_step_kwargs = {}
        if accepts_eta:
            extra_step_kwargs["eta"] = eta

        # check if the scheduler accepts generator
        accepts_generator = "generator" in set(inspect.signature(self.model.noise_scheduler.step).parameters.keys())
        if accepts_generator:
            extra_step_kwargs["generator"] = generator
        return extra_step_kwargs

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.decode_latents
    def decode_latents(self, latents):
        warnings.warn(
            "The decode_latents method is deprecated and will be removed in a future version. Please"
            " use VaeImageProcessor instead",
            FutureWarning,
        )
        latents = 1 / self.model.vae.config.scaling_factor * latents
        image = self.model.vae.decode(latents, return_dict=False)[0]
        image = (image / 2 + 0.5).clamp(0, 1)
        # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
        image = image.cpu().permute(0, 2, 3, 1).float().numpy()
        return image

    def check_inputs(
        self, prompt, callback_steps, negative_prompt=None, prompt_embeds=None, negative_prompt_embeds=None
    ):
        if (callback_steps is None) or (
            callback_steps is not None and (not isinstance(callback_steps, int) or callback_steps <= 0)
        ):
            raise ValueError(
                f"`callback_steps` has to be a positive integer but is {callback_steps} of type"
                f" {type(callback_steps)}."
            )

        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 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)}")

        if negative_prompt is not None and negative_prompt_embeds is not None:
            raise ValueError(
                f"Cannot forward both `negative_prompt`: {negative_prompt} and `negative_prompt_embeds`:"
                f" {negative_prompt_embeds}. Please make sure to only forward one of the two."
            )

        if prompt_embeds is not None and negative_prompt_embeds is not None:
            if prompt_embeds.shape != negative_prompt_embeds.shape:
                raise ValueError(
                    "`prompt_embeds` and `negative_prompt_embeds` must have the same shape when passed directly, but"
                    f" got: `prompt_embeds` {prompt_embeds.shape} != `negative_prompt_embeds`"
                    f" {negative_prompt_embeds.shape}."
                )

    # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_latents
    def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
        shape = (batch_size, num_channels_latents, height // self.model.vae_scale_factor, width // self.model.vae_scale_factor)
        if isinstance(generator, list) and len(generator) != batch_size:
            raise ValueError(
                f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
                f" size of {batch_size}. Make sure the batch size matches the length of the generators."
            )

        if latents is None:
            latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
        else:
            latents = latents.to(device)

        # scale the initial noise by the standard deviation required by the scheduler
        latents = latents * self.model.noise_scheduler.init_noise_sigma
        return latents

    def original_prepare_image_latents(
        self, image, batch_size, num_images_per_prompt, dtype, device, do_classifier_free_guidance, generator=None
    ):
        if not isinstance(image, (torch.Tensor, PIL.Image.Image, list)):
            raise ValueError(
                f"`image` has to be of type `torch.Tensor`, `PIL.Image.Image` or list but is {type(image)}"
            )

        image = image.to(device=device, dtype=dtype)

        batch_size = batch_size * num_images_per_prompt

        if image.shape[1] == 4:
            image_latents = image
        else:
            if isinstance(generator, list) and len(generator) != batch_size:
                raise ValueError(
                    f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
                    f" size of {batch_size}. Make sure the batch size matches the length of the generators."
                )

            if isinstance(generator, list):
                image_latents = [self.model.vae.encode(image[i : i + 1]).latent_dist.mode() for i in range(batch_size)]
                image_latents = torch.cat(image_latents, dim=0)
            else:
                image_latents = self.model.vae.encode(image).latent_dist.mode()

        if batch_size > image_latents.shape[0] and batch_size % image_latents.shape[0] == 0:
            # expand image_latents for batch_size
            deprecation_message = (
                f"You have passed {batch_size} text prompts (`prompt`), but only {image_latents.shape[0]} initial"
                " images (`image`). Initial images are now duplicating to match the number of text prompts. Note"
                " that this behavior is deprecated and will be removed in a version 1.0.0. Please make sure to update"
                " your script to pass as many initial images as text prompts to suppress this warning."
            )
            deprecate("len(prompt) != len(image)", "1.0.0", deprecation_message, standard_warn=False)
            additional_image_per_prompt = batch_size // image_latents.shape[0]
            image_latents = torch.cat([image_latents] * additional_image_per_prompt, dim=0)
        elif batch_size > image_latents.shape[0] and batch_size % image_latents.shape[0] != 0:
            raise ValueError(
                f"Cannot duplicate `image` of batch size {image_latents.shape[0]} to {batch_size} text prompts."
            )
        else:
            image_latents = torch.cat([image_latents], dim=0)

        if do_classifier_free_guidance:
            uncond_image_latents = torch.zeros_like(image_latents)
            image_latents = torch.cat([image_latents, image_latents, uncond_image_latents], dim=0)

        return image_latents
    
    def prepare_image_latents(self, image, batch_size, num_images_per_prompt, dtype, device, do_classifier_free_guidance, generator=None):
        image_latents = self.original_prepare_image_latents(image, batch_size, num_images_per_prompt, dtype, device, do_classifier_free_guidance, generator)
        return repeat(image_latents, 'b c h w -> b c (s h) w', s=self.model.cfg.sequence_length)

    def fancy_get_input_ids(self, prompt, device):
        # textual inversion: procecss multi-vector tokens if necessary
        if isinstance(self, TextualInversionLoaderMixin):
            prompt = self.maybe_convert_prompt(prompt, self.model.tokenizer)

        text_inputs = self.model.tokenizer(
                prompt,
                padding="max_length",
                max_length=self.model.tokenizer.model_max_length,
                truncation=True,
                return_tensors="pt",
            )
        text_input_ids = text_inputs.input_ids
        untruncated_ids = self.model.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.model.tokenizer.batch_decode(
                    untruncated_ids[:, self.model.tokenizer.model_max_length - 1 : -1]
                )
            logger.warning(
                    "The following part of your input was truncated because CLIP can only handle sequences up to"
                    f" {self.model.tokenizer.model_max_length} tokens: {removed_text}"
                )

        if hasattr(self.model.text_encoder.config, "use_attention_mask") and self.model.text_encoder.config.use_attention_mask:
            attention_mask = text_inputs.attention_mask.to(device)
        else:
            attention_mask = None
        text_input_ids = text_input_ids
        return text_input_ids,attention_mask
    
    def encode_prompt_batch(self, 
                        input_ids, 
                        batch_size,
                        device,
                        num_images_per_prompt: int=1,
                        do_classifier_free_guidance: bool=False,
                        negative_prompt=None,
                        prompt_embeds=None,
                        negative_prompt_embeds=None,):
        encoder_hidden_states = self.model.input_ids_to_text_condition(input_ids)
        if self.model.cfg.positional_encoding_type is not None:
            encoder_hidden_states = self.model.apply_step_positional_encoding(encoder_hidden_states)
        prompt_embeds = encoder_hidden_states
        prompt_embeds = prompt_embeds.to(dtype=self.model.text_encoder.dtype, device=device)

        bs_embed, seq_len, _ = prompt_embeds.shape
        # duplicate text embeddings 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(bs_embed * num_images_per_prompt, seq_len, -1)
        
        if do_classifier_free_guidance:
            if negative_prompt_embeds is None:
                negative_prompt_embeds = self.model.get_null_conditioning()
                negative_prompt_embeds = repeat(negative_prompt_embeds, 'o t l -> (b o) t l', b=batch_size) #, o=1
            # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
            seq_len = negative_prompt_embeds.shape[1]

            negative_prompt_embeds = negative_prompt_embeds.to(dtype=self.model.text_encoder.dtype, device=device)

            negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
            negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)

            # For classifier free guidance, we need to do two forward passes.
            # Here we concatenate the unconditional and text embeddings into a single batch
            # to avoid doing two forward passes
            # pix2pix has two  negative embeddings, and unlike in other pipelines latents are ordered [prompt_embeds, negative_prompt_embeds, negative_prompt_embeds]
            prompt_embeds = torch.cat([prompt_embeds, negative_prompt_embeds, negative_prompt_embeds])
        return prompt_embeds