flux pipeline

app.py

@@ -8,7 +8,7 @@ from sid import SiDFluxPipeline, SiDSD3Pipeline, SiDSanaPipeline
 import torch
 
 device = "cuda" if torch.cuda.is_available() else "cpu"
-
+torch_dtype = torch.float16
 
 MODEL_OPTIONS = {
     "SiD-Flow-SD3-medium": "YGu1998/SiD-Flow-SD3-medium",
@@ -31,13 +31,13 @@ def load_model(model_choice):
     model_repo_id = MODEL_OPTIONS[model_choice]
     time_scale = 1000.0
     if "Sana" in model_choice:
-        pipe = SiDSanaPipeline.from_pretrained(model_repo_id, torch_dtype=torch.float16)
+        pipe = SiDSanaPipeline.from_pretrained(model_repo_id, torch_dtype=torch_dtype)
         if "Sprint" in model_choice:
             time_scale = 1.0
     elif "SD3" in model_choice:
-        pipe = SiDSD3Pipeline.from_pretrained(model_repo_id, torch_dtype=torch.float16)
+        pipe = SiDSD3Pipeline.from_pretrained(model_repo_id, torch_dtype=torch_dtype)
     elif "Flux" in model_choice:
-        pipe = SiDFluxPipeline.from_pretrained(model_repo_id, torch_dtype=torch.float16)
+        pipe = SiDFluxPipeline.from_pretrained(model_repo_id, torch_dtype=torch_dtype)
     else:
         raise ValueError(f"Unknown model type for: {model_choice}")
     pipe = pipe.to(device)

sid/pipeline_sid_flux.py

@@ -27,7 +27,12 @@ from transformers import (
 )
 
 from diffusers.image_processor import PipelineImageInput, VaeImageProcessor
-from diffusers.loaders import FluxIPAdapterMixin, FluxLoraLoaderMixin, FromSingleFileMixin, TextualInversionLoaderMixin
+from diffusers.loaders import (
+    FluxIPAdapterMixin,
+    FluxLoraLoaderMixin,
+    FromSingleFileMixin,
+    TextualInversionLoaderMixin,
+)
 from diffusers.models import AutoencoderKL, FluxTransformer2DModel
 from diffusers.schedulers import FlowMatchEulerDiscreteScheduler
 from diffusers.utils import (
@@ -53,22 +58,6 @@ else:
 
 logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
 
-EXAMPLE_DOC_STRING = """
-    Examples:
-        ```py
-        >>> import torch
-        >>> from diffusers import FluxPipeline
-
-        >>> pipe = FluxPipeline.from_pretrained("black-forest-labs/FLUX.1-schnell", torch_dtype=torch.bfloat16)
-        >>> pipe.to("cuda")
-        >>> prompt = "A cat holding a sign that says hello world"
-        >>> # Depending on the variant being used, the pipeline call will slightly vary.
-        >>> # Refer to the pipeline documentation for more details.
-        >>> image = pipe(prompt, num_inference_steps=4, guidance_scale=0.0).images[0]
-        >>> image.save("flux.png")
-        ```
-"""
-
 
 def calculate_shift(
     image_seq_len,
@@ -116,9 +105,13 @@ def retrieve_timesteps(
         second element is the number of inference steps.
     """
     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")
+        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())
+        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"
@@ -128,7 +121,9 @@ def retrieve_timesteps(
         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())
+        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"
@@ -176,7 +171,9 @@ class SiDFluxPipeline(
             [T5TokenizerFast](https://huggingface.co/docs/transformers/en/model_doc/t5#transformers.T5TokenizerFast).
     """
 
-    model_cpu_offload_seq = "text_encoder->text_encoder_2->image_encoder->transformer->vae"
+    model_cpu_offload_seq = (
+        "text_encoder->text_encoder_2->image_encoder->transformer->vae"
+    )
     _optional_components = ["image_encoder", "feature_extractor"]
     _callback_tensor_inputs = ["latents", "prompt_embeds"]
 
@@ -205,12 +202,20 @@ class SiDFluxPipeline(
             image_encoder=image_encoder,
             feature_extractor=feature_extractor,
         )
-        self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
+        self.vae_scale_factor = (
+            2 ** (len(self.vae.config.block_out_channels) - 1)
+            if getattr(self, "vae", None)
+            else 8
+        )
         # Flux latents are turned into 2x2 patches and packed. This means the latent width and height has to be divisible
         # by the patch size. So the vae scale factor is multiplied by the patch size to account for this
-        self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor * 2)
+        self.image_processor = VaeImageProcessor(
+            vae_scale_factor=self.vae_scale_factor * 2
+        )
         self.tokenizer_max_length = (
-            self.tokenizer.model_max_length if hasattr(self, "tokenizer") and self.tokenizer is not None else 77
+            self.tokenizer.model_max_length
+            if hasattr(self, "tokenizer") and self.tokenizer is not None
+            else 77
         )
         self.default_sample_size = 128
 
@@ -241,16 +246,24 @@ class SiDFluxPipeline(
             return_tensors="pt",
         )
         text_input_ids = text_inputs.input_ids
-        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").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])
+        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]
+        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)
@@ -259,7 +272,9 @@ class SiDFluxPipeline(
 
         # 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)
+        prompt_embeds = prompt_embeds.view(
+            batch_size * num_images_per_prompt, seq_len, -1
+        )
 
         return prompt_embeds
 
@@ -288,14 +303,22 @@ class SiDFluxPipeline(
         )
 
         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])
+        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)
+        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
@@ -381,7 +404,11 @@ class SiDFluxPipeline(
                 # 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
+        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
@@ -405,19 +432,27 @@ class SiDFluxPipeline(
             if not isinstance(ip_adapter_image, list):
                 ip_adapter_image = [ip_adapter_image]
 
-            if len(ip_adapter_image) != self.transformer.encoder_hid_proj.num_ip_adapters:
+            if (
+                len(ip_adapter_image)
+                != self.transformer.encoder_hid_proj.num_ip_adapters
+            ):
                 raise ValueError(
                     f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {self.transformer.encoder_hid_proj.num_ip_adapters} IP Adapters."
                 )
 
             for single_ip_adapter_image in ip_adapter_image:
-                single_image_embeds = self.encode_image(single_ip_adapter_image, device, 1)
+                single_image_embeds = self.encode_image(
+                    single_ip_adapter_image, device, 1
+                )
                 image_embeds.append(single_image_embeds[None, :])
         else:
             if not isinstance(ip_adapter_image_embeds, list):
                 ip_adapter_image_embeds = [ip_adapter_image_embeds]
 
-            if len(ip_adapter_image_embeds) != self.transformer.encoder_hid_proj.num_ip_adapters:
+            if (
+                len(ip_adapter_image_embeds)
+                != self.transformer.encoder_hid_proj.num_ip_adapters
+            ):
                 raise ValueError(
                     f"`ip_adapter_image_embeds` must have same length as the number of IP Adapters. Got {len(ip_adapter_image_embeds)} image embeds and {self.transformer.encoder_hid_proj.num_ip_adapters} IP Adapters."
                 )
@@ -427,7 +462,9 @@ class SiDFluxPipeline(
 
         ip_adapter_image_embeds = []
         for single_image_embeds in image_embeds:
-            single_image_embeds = torch.cat([single_image_embeds] * num_images_per_prompt, dim=0)
+            single_image_embeds = torch.cat(
+                [single_image_embeds] * num_images_per_prompt, dim=0
+            )
             single_image_embeds = single_image_embeds.to(device=device)
             ip_adapter_image_embeds.append(single_image_embeds)
 
@@ -448,13 +485,17 @@ class SiDFluxPipeline(
         callback_on_step_end_tensor_inputs=None,
         max_sequence_length=None,
     ):
-        if height % (self.vae_scale_factor * 2) != 0 or width % (self.vae_scale_factor * 2) != 0:
+        if (
+            height % (self.vae_scale_factor * 2) != 0
+            or width % (self.vae_scale_factor * 2) != 0
+        ):
             logger.warning(
                 f"`height` and `width` have to be divisible by {self.vae_scale_factor * 2} but are {height} and {width}. Dimensions will be resized accordingly"
             )
 
         if callback_on_step_end_tensor_inputs is not None and not all(
-            k in self._callback_tensor_inputs for k in callback_on_step_end_tensor_inputs
+            k in self._callback_tensor_inputs
+            for k in callback_on_step_end_tensor_inputs
         ):
             raise ValueError(
                 f"`callback_on_step_end_tensor_inputs` has to be in {self._callback_tensor_inputs}, but found {[k for k in callback_on_step_end_tensor_inputs if k not in self._callback_tensor_inputs]}"
@@ -474,10 +515,18 @@ class SiDFluxPipeline(
             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)}")
+        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 negative_prompt is not None and negative_prompt_embeds is not None:
             raise ValueError(
@@ -500,15 +549,23 @@ class SiDFluxPipeline(
             )
 
         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}")
+            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, width, 3)
-        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height)[:, None]
-        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width)[None, :]
+        latent_image_ids[..., 1] = (
+            latent_image_ids[..., 1] + torch.arange(height)[:, None]
+        )
+        latent_image_ids[..., 2] = (
+            latent_image_ids[..., 2] + torch.arange(width)[None, :]
+        )
 
-        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
+        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
@@ -518,9 +575,13 @@ class SiDFluxPipeline(
 
     @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.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)
+        latents = latents.reshape(
+            batch_size, (height // 2) * (width // 2), num_channels_latents * 4
+        )
 
         return latents
 
@@ -588,7 +649,9 @@ class SiDFluxPipeline(
         shape = (batch_size, num_channels_latents, height, width)
 
         if latents is not None:
-            latent_image_ids = self._prepare_latent_image_ids(batch_size, height // 2, width // 2, device, dtype)
+            latent_image_ids = self._prepare_latent_image_ids(
+                batch_size, height // 2, width // 2, device, dtype
+            )
             return latents.to(device=device, dtype=dtype), latent_image_ids
 
         if isinstance(generator, list) and len(generator) != batch_size:
@@ -598,9 +661,13 @@ class SiDFluxPipeline(
             )
 
         latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
-        latents = self._pack_latents(latents, batch_size, num_channels_latents, height, width)
+        latents = self._pack_latents(
+            latents, batch_size, num_channels_latents, height, width
+        )
 
-        latent_image_ids = self._prepare_latent_image_ids(batch_size, height // 2, width // 2, device, dtype)
+        latent_image_ids = self._prepare_latent_image_ids(
+            batch_size, height // 2, width // 2, device, dtype
+        )
 
         return latents, latent_image_ids
 
@@ -625,135 +692,29 @@ class SiDFluxPipeline(
         return self._interrupt
 
     @torch.no_grad()
-    @replace_example_docstring(EXAMPLE_DOC_STRING)
     def __call__(
         self,
         prompt: Union[str, List[str]] = None,
         prompt_2: Optional[Union[str, List[str]]] = None,
-        negative_prompt: Union[str, List[str]] = None,
-        negative_prompt_2: Optional[Union[str, List[str]]] = None,
         true_cfg_scale: float = 1.0,
         height: Optional[int] = None,
         width: Optional[int] = None,
         num_inference_steps: int = 28,
         sigmas: Optional[List[float]] = None,
-        guidance_scale: float = 3.5,
+        guidance_scale: float = 1,
         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,
-        ip_adapter_image: Optional[PipelineImageInput] = None,
-        ip_adapter_image_embeds: Optional[List[torch.Tensor]] = None,
-        negative_ip_adapter_image: Optional[PipelineImageInput] = None,
-        negative_ip_adapter_image_embeds: Optional[List[torch.Tensor]] = None,
-        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
-        negative_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,
+        noise_type: str = "fresh",  # 'fresh', 'ddim', 'fixed'
     ):
-        r"""
-        Function invoked when calling the pipeline for generation.
-
-        Args:
-            prompt (`str` or `List[str]`, *optional*):
-                The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`.
-                instead.
-            prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts to be sent to `tokenizer_2` and `text_encoder_2`. If not defined, `prompt` is
-                will be used instead.
-            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 `true_cfg_scale` is
-                not greater than `1`).
-            negative_prompt_2 (`str` or `List[str]`, *optional*):
-                The prompt or prompts not to guide the image generation to be sent to `tokenizer_2` and
-                `text_encoder_2`. If not defined, `negative_prompt` is used in all the text-encoders.
-            true_cfg_scale (`float`, *optional*, defaults to 1.0):
-                True classifier-free guidance (guidance scale) is enabled when `true_cfg_scale` > 1 and
-                `negative_prompt` is provided.
-            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.
-            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):
-                Embedded guiddance scale is enabled by setting `guidance_scale` > 1. Higher `guidance_scale` encourages
-                a model to generate images more aligned with `prompt` at the expense of lower image quality.
-
-                Guidance-distilled models approximates true classifer-free guidance for `guidance_scale` > 1. Refer to
-                the [paper](https://huggingface.co/papers/2210.03142) to learn more.
-            num_images_per_prompt (`int`, *optional*, defaults to 1):
-                The number of images to generate per prompt.
-            generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
-                One or a list of [torch generator(s)](https://pytorch.org/docs/stable/generated/torch.Generator.html)
-                to make generation deterministic.
-            latents (`torch.FloatTensor`, *optional*):
-                Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
-                generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
-                tensor will be generated by sampling using the supplied random `generator`.
-            prompt_embeds (`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.
-            pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
-                If not provided, pooled text embeddings will be generated from `prompt` input argument.
-            ip_adapter_image: (`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.
-            ip_adapter_image_embeds (`List[torch.Tensor]`, *optional*):
-                Pre-generated image embeddings for IP-Adapter. It should be a list of length same as number of
-                IP-adapters. Each element should be a tensor of shape `(batch_size, num_images, emb_dim)`. If not
-                provided, embeddings are computed from the `ip_adapter_image` input argument.
-            negative_ip_adapter_image:
-                (`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.
-            negative_ip_adapter_image_embeds (`List[torch.Tensor]`, *optional*):
-                Pre-generated image embeddings for IP-Adapter. It should be a list of length same as number of
-                IP-adapters. Each element should be a tensor of shape `(batch_size, num_images, emb_dim)`. If not
-                provided, embeddings are computed from the `ip_adapter_image` 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.
-            negative_pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
-                Pre-generated negative pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
-                weighting. If not provided, pooled 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.flux.FluxPipelineOutput`] instead of a plain tuple.
-            joint_attention_kwargs (`dict`, *optional*):
-                A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
-                `self.processor` in
-                [diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
-            callback_on_step_end (`Callable`, *optional*):
-                A function that calls at the end of each denoising steps during the inference. The function is called
-                with the following arguments: `callback_on_step_end(self: DiffusionPipeline, step: int, timestep: int,
-                callback_kwargs: Dict)`. `callback_kwargs` will include a list of all tensors as specified by
-                `callback_on_step_end_tensor_inputs`.
-            callback_on_step_end_tensor_inputs (`List`, *optional*):
-                The list of tensor inputs for the `callback_on_step_end` function. The tensors specified in the list
-                will be passed as `callback_kwargs` argument. You will only be able to include variables listed in the
-                `._callback_tensor_inputs` attribute of your pipeline class.
-            max_sequence_length (`int` defaults to 512): Maximum sequence length to use with the `prompt`.
-
-        Examples:
-
-        Returns:
-            [`~pipelines.flux.FluxPipelineOutput`] or `tuple`: [`~pipelines.flux.FluxPipelineOutput`] if `return_dict`
-            is True, otherwise a `tuple`. When returning a tuple, the first element is a list with the generated
-            images.
-        """
 
         height = height or self.default_sample_size * self.vae_scale_factor
         width = width or self.default_sample_size * self.vae_scale_factor
@@ -764,12 +725,8 @@ class SiDFluxPipeline(
             prompt_2,
             height,
             width,
-            negative_prompt=negative_prompt,
-            negative_prompt_2=negative_prompt_2,
             prompt_embeds=prompt_embeds,
-            negative_prompt_embeds=negative_prompt_embeds,
             pooled_prompt_embeds=pooled_prompt_embeds,
-            negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,
             callback_on_step_end_tensor_inputs=callback_on_step_end_tensor_inputs,
             max_sequence_length=max_sequence_length,
         )
@@ -789,13 +746,6 @@ class SiDFluxPipeline(
 
         device = self._execution_device
 
-        lora_scale = (
-            self.joint_attention_kwargs.get("scale", None) if self.joint_attention_kwargs is not None else None
-        )
-        has_neg_prompt = negative_prompt is not None or (
-            negative_prompt_embeds is not None and negative_pooled_prompt_embeds is not None
-        )
-        do_true_cfg = true_cfg_scale > 1 and has_neg_prompt
         (
             prompt_embeds,
             pooled_prompt_embeds,
@@ -808,23 +758,7 @@ class SiDFluxPipeline(
             device=device,
             num_images_per_prompt=num_images_per_prompt,
             max_sequence_length=max_sequence_length,
-            lora_scale=lora_scale,
         )
-        if do_true_cfg:
-            (
-                negative_prompt_embeds,
-                negative_pooled_prompt_embeds,
-                negative_text_ids,
-            ) = self.encode_prompt(
-                prompt=negative_prompt,
-                prompt_2=negative_prompt_2,
-                prompt_embeds=negative_prompt_embeds,
-                pooled_prompt_embeds=negative_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
@@ -839,147 +773,82 @@ class SiDFluxPipeline(
             latents,
         )
 
-        # 5. Prepare timesteps
-        sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps) if sigmas is None else sigmas
-        if hasattr(self.scheduler.config, "use_flow_sigmas") and self.scheduler.config.use_flow_sigmas:
-            sigmas = None
-        image_seq_len = latents.shape[1]
-        mu = calculate_shift(
-            image_seq_len,
-            self.scheduler.config.get("base_image_seq_len", 256),
-            self.scheduler.config.get("max_image_seq_len", 4096),
-            self.scheduler.config.get("base_shift", 0.5),
-            self.scheduler.config.get("max_shift", 1.15),
-        )
-        timesteps, num_inference_steps = retrieve_timesteps(
-            self.scheduler,
-            num_inference_steps,
-            device,
-            sigmas=sigmas,
-            mu=mu,
-        )
-        num_warmup_steps = max(len(timesteps) - num_inference_steps * self.scheduler.order, 0)
-        self._num_timesteps = len(timesteps)
+        # Denoising loop
+        D_x = torch.zeros_like(latents).to(latents.device)
+        initial_latents = latents.clone() if noise_type == "fixed" else None
 
-        # 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
-
-        if (ip_adapter_image is not None or ip_adapter_image_embeds is not None) and (
-            negative_ip_adapter_image is None and negative_ip_adapter_image_embeds is None
-        ):
-            negative_ip_adapter_image = np.zeros((width, height, 3), dtype=np.uint8)
-            negative_ip_adapter_image = [negative_ip_adapter_image] * self.transformer.encoder_hid_proj.num_ip_adapters
+        for i in range(num_inference_steps):
 
-        elif (ip_adapter_image is None and ip_adapter_image_embeds is None) and (
-            negative_ip_adapter_image is not None or negative_ip_adapter_image_embeds is not None
-        ):
-            ip_adapter_image = np.zeros((width, height, 3), dtype=np.uint8)
-            ip_adapter_image = [ip_adapter_image] * self.transformer.encoder_hid_proj.num_ip_adapters
-
-        if self.joint_attention_kwargs is None:
-            self._joint_attention_kwargs = {}
-
-        image_embeds = None
-        negative_image_embeds = None
-        if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
-            image_embeds = self.prepare_ip_adapter_image_embeds(
-                ip_adapter_image,
-                ip_adapter_image_embeds,
-                device,
-                batch_size * num_images_per_prompt,
+            if noise_type == "fresh":
+                noise = (
+                    latents if i == 0 else torch.randn_like(latents).to(latents.device)
+                )
+            elif noise_type == "ddim":
+                noise = (
+                    latents if i == 0 else ((latents - (1.0 - t) * D_x) / t).detach()
+                )
+            elif noise_type == "fixed":
+                noise = initial_latents  # Use the initial, unmodified latents
+            else:
+                raise ValueError(f"Unknown noise_type: {noise_type}")
+
+            with torch.no_grad():
+                # Compute timestep t for current denoising step, normalized to [0, 1]
+                scalar_t = 999.0 * (1.0 - float(i) / float(num_inference_steps - 1))
+                t_val = scalar_t / 999.0
+                t = torch.full(
+                    (latents.shape[0],),
+                    t_val,
+                    device=latents.device,
+                    dtype=latents.dtype,
+                )
+                if t.numel() > 1:
+                    t = t.view(-1, 1, 1, 1)
+
+            latents = (1.0 - t) * D_x + t * noise
+            latent_image_ids = self._prepare_latent_image_ids(
+                latents.shape[0],
+                latents.shape[2] // 2,
+                latents.shape[3] // 2,
+                latents.device,
+                latents.dtype,
             )
-        if negative_ip_adapter_image is not None or negative_ip_adapter_image_embeds is not None:
-            negative_image_embeds = self.prepare_ip_adapter_image_embeds(
-                negative_ip_adapter_image,
-                negative_ip_adapter_image_embeds,
-                device,
-                batch_size * num_images_per_prompt,
+            packed_latents = self._pack_latents(
+                latents,
+                batch_size=latents.shape[0],
+                num_channels_latents=latents.shape[1],
+                height=latents.shape[2],
+                width=latents.shape[3],
             )
 
-        # 6. Denoising loop
-        # We set the index here to remove DtoH sync, helpful especially during compilation.
-        # Check out more details here: https://github.com/huggingface/diffusers/pull/11696
-        self.scheduler.set_begin_index(0)
-        with self.progress_bar(total=num_inference_steps) as progress_bar:
-            for i, t in enumerate(timesteps):
-                if self.interrupt:
-                    continue
-
-                self._current_timestep = t
-                if image_embeds is not None:
-                    self._joint_attention_kwargs["ip_adapter_image_embeds"] = image_embeds
-                # broadcast to batch dimension in a way that's compatible with ONNX/Core ML
-                timestep = t.expand(latents.shape[0]).to(latents.dtype)
-
-                with self.transformer.cache_context("cond"):
-                    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]
-
-                if do_true_cfg:
-                    if negative_image_embeds is not None:
-                        self._joint_attention_kwargs["ip_adapter_image_embeds"] = negative_image_embeds
-
-                    with self.transformer.cache_context("uncond"):
-                        neg_noise_pred = self.transformer(
-                            hidden_states=latents,
-                            timestep=timestep / 1000,
-                            guidance=guidance,
-                            pooled_projections=negative_pooled_prompt_embeds,
-                            encoder_hidden_states=negative_prompt_embeds,
-                            txt_ids=negative_text_ids,
-                            img_ids=latent_image_ids,
-                            joint_attention_kwargs=self.joint_attention_kwargs,
-                            return_dict=False,
-                        )[0]
-                    noise_pred = neg_noise_pred + true_cfg_scale * (noise_pred - neg_noise_pred)
-
-                # 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()
+            guidance = torch.tensor([guidance_scale], device=device)
+            guidance = guidance.expand(latents.shape[0])
 
-        self._current_timestep = None
+            flow_pred = self.transformer(
+                hidden_states=packed_latents,
+                # YiYi notes: divide it by 1000 for now because we scale it by 1000 in the transformer model (we should not keep it but I want to keep the inputs same for the model for testing)
+                timestep=t.view(-1),  # timesteps / 1000.0,
+                guidance=guidance,
+                pooled_projections=pooled_prompt_embeds,
+                encoder_hidden_states=prompt_embeds,
+                txt_ids=text_ids,
+                img_ids=latent_image_ids,
+                return_dict=False,
+            )[0]
+
+            flow_pred = self._unpack_latents(
+                flow_pred,
+                height=height * self.vae_scale_factor,
+                width=width * self.vae_scale_factor,
+                vae_scale_factor=self.vae_scale_factor,
+            )
+            D_x = latents - t.view(-1, 1, 1, 1) * flow_pred
 
-        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, return_dict=False)[0]
-            image = self.image_processor.postprocess(image, output_type=output_type)
+        latents = (
+            latents / self.vae.config.scaling_factor
+        ) + self.vae.config.shift_factor
+        image = self.vae.decode(latents, return_dict=False)[0]
+        image = self.image_processor.postprocess(image, output_type=output_type)
 
         # Offload all models
         self.maybe_free_model_hooks()
@@ -987,4 +856,4 @@ class SiDFluxPipeline(
         if not return_dict:
             return (image,)
 
-        return FluxPipelineOutput(images=image)
+        return SiDPipelineOutput(images=image)

sid/pipeline_sid_sana.py

@@ -159,9 +159,13 @@ def retrieve_timesteps(
         second element is the number of inference steps.
     """
     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")
+        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())
+        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"
@@ -171,7 +175,9 @@ def retrieve_timesteps(
         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())
+        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"
@@ -209,7 +215,11 @@ class SiDSanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         super().__init__()
 
         self.register_modules(
-            tokenizer=tokenizer, text_encoder=text_encoder, vae=vae, transformer=transformer, scheduler=scheduler
+            tokenizer=tokenizer,
+            text_encoder=text_encoder,
+            vae=vae,
+            transformer=transformer,
+            scheduler=scheduler,
         )
 
         self.vae_scale_factor = (
@@ -217,7 +227,9 @@ class SiDSanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
             if hasattr(self, "vae") and self.vae is not None
             else 32
         )
-        self.image_processor = PixArtImageProcessor(vae_scale_factor=self.vae_scale_factor)
+        self.image_processor = PixArtImageProcessor(
+            vae_scale_factor=self.vae_scale_factor
+        )
 
     def enable_vae_slicing(self):
         r"""
@@ -301,7 +313,9 @@ class SiDSanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         prompt_attention_mask = text_inputs.attention_mask
         prompt_attention_mask = prompt_attention_mask.to(device)
 
-        prompt_embeds = self.text_encoder(text_input_ids.to(device), attention_mask=prompt_attention_mask)
+        prompt_embeds = self.text_encoder(
+            text_input_ids.to(device), attention_mask=prompt_attention_mask
+        )
         prompt_embeds = prompt_embeds[0].to(dtype=dtype, device=device)
 
         return prompt_embeds, prompt_attention_mask
@@ -398,33 +412,51 @@ class SiDSanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         bs_embed, 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(bs_embed * num_images_per_prompt, seq_len, -1)
+        prompt_embeds = prompt_embeds.view(
+            bs_embed * num_images_per_prompt, seq_len, -1
+        )
         prompt_attention_mask = prompt_attention_mask.view(bs_embed, -1)
         prompt_attention_mask = prompt_attention_mask.repeat(num_images_per_prompt, 1)
 
         # get unconditional embeddings for classifier free guidance
         if do_classifier_free_guidance and negative_prompt_embeds is None:
-            negative_prompt = [negative_prompt] * batch_size if isinstance(negative_prompt, str) else negative_prompt
-            negative_prompt_embeds, negative_prompt_attention_mask = self._get_gemma_prompt_embeds(
-                prompt=negative_prompt,
-                device=device,
-                dtype=dtype,
-                clean_caption=clean_caption,
-                max_sequence_length=max_sequence_length,
-                complex_human_instruction=False,
+            negative_prompt = (
+                [negative_prompt] * batch_size
+                if isinstance(negative_prompt, str)
+                else negative_prompt
+            )
+            negative_prompt_embeds, negative_prompt_attention_mask = (
+                self._get_gemma_prompt_embeds(
+                    prompt=negative_prompt,
+                    device=device,
+                    dtype=dtype,
+                    clean_caption=clean_caption,
+                    max_sequence_length=max_sequence_length,
+                    complex_human_instruction=False,
+                )
             )
 
         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=dtype, device=device)
+            negative_prompt_embeds = negative_prompt_embeds.to(
+                dtype=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)
+            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
+            )
 
-            negative_prompt_attention_mask = negative_prompt_attention_mask.view(bs_embed, -1)
-            negative_prompt_attention_mask = negative_prompt_attention_mask.repeat(num_images_per_prompt, 1)
+            negative_prompt_attention_mask = negative_prompt_attention_mask.view(
+                bs_embed, -1
+            )
+            negative_prompt_attention_mask = negative_prompt_attention_mask.repeat(
+                num_images_per_prompt, 1
+            )
         else:
             negative_prompt_embeds = None
             negative_prompt_attention_mask = None
@@ -434,7 +466,12 @@ class SiDSanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
                 # Retrieve the original scale by scaling back the LoRA layers
                 unscale_lora_layers(self.text_encoder, lora_scale)
 
-        return prompt_embeds, prompt_attention_mask, negative_prompt_embeds, negative_prompt_attention_mask
+        return (
+            prompt_embeds,
+            prompt_attention_mask,
+            negative_prompt_embeds,
+            negative_prompt_attention_mask,
+        )
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs
     def prepare_extra_step_kwargs(self, generator, eta):
@@ -443,13 +480,17 @@ class SiDSanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         # eta corresponds to η in DDIM paper: https://huggingface.co/papers/2010.02502
         # and should be between [0, 1]
 
-        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
+        accepts_eta = "eta" in set(
+            inspect.signature(self.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.scheduler.step).parameters.keys())
+        accepts_generator = "generator" in set(
+            inspect.signature(self.scheduler.step).parameters.keys()
+        )
         if accepts_generator:
             extra_step_kwargs["generator"] = generator
         return extra_step_kwargs
@@ -467,10 +508,13 @@ class SiDSanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         negative_prompt_attention_mask=None,
     ):
         if height % 32 != 0 or width % 32 != 0:
-            raise ValueError(f"`height` and `width` have to be divisible by 32 but are {height} and {width}.")
+            raise ValueError(
+                f"`height` and `width` have to be divisible by 32 but are {height} and {width}."
+            )
 
         if callback_on_step_end_tensor_inputs is not None and not all(
-            k in self._callback_tensor_inputs for k in callback_on_step_end_tensor_inputs
+            k in self._callback_tensor_inputs
+            for k in callback_on_step_end_tensor_inputs
         ):
             raise ValueError(
                 f"`callback_on_step_end_tensor_inputs` has to be in {self._callback_tensor_inputs}, but found {[k for k in callback_on_step_end_tensor_inputs if k not in self._callback_tensor_inputs]}"
@@ -485,8 +529,12 @@ class SiDSanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
             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 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 prompt is not None and negative_prompt_embeds is not None:
             raise ValueError(
@@ -501,10 +549,17 @@ class SiDSanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
             )
 
         if prompt_embeds is not None and prompt_attention_mask is None:
-            raise ValueError("Must provide `prompt_attention_mask` when specifying `prompt_embeds`.")
+            raise ValueError(
+                "Must provide `prompt_attention_mask` when specifying `prompt_embeds`."
+            )
 
-        if negative_prompt_embeds is not None and negative_prompt_attention_mask is None:
-            raise ValueError("Must provide `negative_prompt_attention_mask` when specifying `negative_prompt_embeds`.")
+        if (
+            negative_prompt_embeds is not None
+            and negative_prompt_attention_mask is None
+        ):
+            raise ValueError(
+                "Must provide `negative_prompt_attention_mask` when specifying `negative_prompt_embeds`."
+            )
 
         if prompt_embeds is not None and negative_prompt_embeds is not None:
             if prompt_embeds.shape != negative_prompt_embeds.shape:
@@ -523,12 +578,16 @@ class SiDSanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
     # Copied from diffusers.pipelines.deepfloyd_if.pipeline_if.IFPipeline._text_preprocessing
     def _text_preprocessing(self, text, clean_caption=False):
         if clean_caption and not is_bs4_available():
-            logger.warning(BACKENDS_MAPPING["bs4"][-1].format("Setting `clean_caption=True`"))
+            logger.warning(
+                BACKENDS_MAPPING["bs4"][-1].format("Setting `clean_caption=True`")
+            )
             logger.warning("Setting `clean_caption` to False...")
             clean_caption = False
 
         if clean_caption and not is_ftfy_available():
-            logger.warning(BACKENDS_MAPPING["ftfy"][-1].format("Setting `clean_caption=True`"))
+            logger.warning(
+                BACKENDS_MAPPING["ftfy"][-1].format("Setting `clean_caption=True`")
+            )
             logger.warning("Setting `clean_caption` to False...")
             clean_caption = False
 
@@ -616,13 +675,17 @@ class SiDSanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         # "123456.."
         caption = re.sub(r"\b\d{6,}\b", "", caption)
         # filenames:
-        caption = re.sub(r"[\S]+\.(?:png|jpg|jpeg|bmp|webp|eps|pdf|apk|mp4)", "", caption)
+        caption = re.sub(
+            r"[\S]+\.(?:png|jpg|jpeg|bmp|webp|eps|pdf|apk|mp4)", "", caption
+        )
 
         #
         caption = re.sub(r"[\"\']{2,}", r'"', caption)  # """AUSVERKAUFT"""
         caption = re.sub(r"[\.]{2,}", r" ", caption)  # """AUSVERKAUFT"""
 
-        caption = re.sub(self.bad_punct_regex, r" ", caption)  # ***AUSVERKAUFT***, #AUSVERKAUFT
+        caption = re.sub(
+            self.bad_punct_regex, r" ", caption
+        )  # ***AUSVERKAUFT***, #AUSVERKAUFT
         caption = re.sub(r"\s+\.\s+", r" ", caption)  # " . "
 
         # this-is-my-cute-cat / this_is_my_cute_cat
@@ -640,10 +703,14 @@ class SiDSanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         caption = re.sub(r"(worldwide\s+)?(free\s+)?shipping", "", caption)
         caption = re.sub(r"(free\s)?download(\sfree)?", "", caption)
         caption = re.sub(r"\bclick\b\s(?:for|on)\s\w+", "", caption)
-        caption = re.sub(r"\b(?:png|jpg|jpeg|bmp|webp|eps|pdf|apk|mp4)(\simage[s]?)?", "", caption)
+        caption = re.sub(
+            r"\b(?:png|jpg|jpeg|bmp|webp|eps|pdf|apk|mp4)(\simage[s]?)?", "", caption
+        )
         caption = re.sub(r"\bpage\s+\d+\b", "", caption)
 
-        caption = re.sub(r"\b\d*[a-zA-Z]+\d+[a-zA-Z]+\d+[a-zA-Z\d]*\b", r" ", caption)  # j2d1a2a...
+        caption = re.sub(
+            r"\b\d*[a-zA-Z]+\d+[a-zA-Z]+\d+[a-zA-Z\d]*\b", r" ", caption
+        )  # j2d1a2a...
 
         caption = re.sub(r"\b\d+\.?\d*[xх×]\d+\.?\d*\b", "", caption)
 
@@ -660,7 +727,17 @@ class SiDSanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
 
         return caption.strip()
 
-    def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
+    def prepare_latents(
+        self,
+        batch_size,
+        num_channels_latents,
+        height,
+        width,
+        dtype,
+        device,
+        generator,
+        latents=None,
+    ):
         if latents is not None:
             return latents.to(device=device, dtype=dtype)
 
@@ -733,8 +810,10 @@ class SiDSanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
             else:
                 raise ValueError("Invalid sample size")
             orig_height, orig_width = height, width
-            height, width = self.image_processor.classify_height_width_bin(height, width, ratios=aspect_ratio_bin)
-        
+            height, width = self.image_processor.classify_height_width_bin(
+                height, width, ratios=aspect_ratio_bin
+            )
+
         if isinstance(callback_on_step_end, (PipelineCallback, MultiPipelineCallbacks)):
             callback_on_step_end_tensor_inputs = callback_on_step_end.tensor_inputs
 
@@ -764,7 +843,8 @@ class SiDSanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
         (
             prompt_embeds,
             prompt_attention_mask,
-            _, _,
+            _,
+            _,
         ) = self.encode_prompt(
             prompt,
             prompt_embeds=prompt_embeds,
@@ -840,7 +920,9 @@ class SiDSanaPipeline(DiffusionPipeline, SanaLoraLoaderMixin):
             return_dict=False,
         )[0]
         if use_resolution_binning:
-            image = self.image_processor.resize_and_crop_tensor(image, orig_height, orig_width)
+            image = self.image_processor.resize_and_crop_tensor(
+                image, orig_height, orig_width
+            )
         image = self.image_processor.postprocess(image, output_type=output_type)
         # Offload all models
         self.maybe_free_model_hooks()

sid/pipeline_sid_sd3.py

@@ -749,16 +749,16 @@ class SiDSD3Pipeline(
         # Initialize D_x
         D_x = torch.zeros_like(latents).to(latents.device)
         # Use fixed noise for now (can be extended as needed)
-        initial_latents = latents.clone()
+        initial_latents = latents.clone() if noise_type == 'fixed' else None
         for i in range(num_inference_steps):
             if noise_type == "fresh":
                 noise = (
                     latents if i == 0 else torch.randn_like(latents).to(latents.device)
                 )
             elif noise_type == "ddim":
-                noise = (
-                    latents if i == 0 else ((latents - (1.0 - t) * D_x) / t).detach()
-                )
+                    noise = (
+                        latents if i == 0 else ((latents - (1.0 - t) * D_x) / t).detach()
+                    )
             elif noise_type == "fixed":
                 noise = initial_latents  # Use the initial, unmodified latents
             else: