diffusers 0.27.2__py3-none-any.whl → 0.28.0__py3-none-any.whl

diffusers/pipelines/animatediff/pipeline_animatediff_video2video.py

@@ -15,11 +15,10 @@
 import inspect
 from typing import Any, Callable, Dict, List, Optional, Union
 
-import numpy as np
 import torch
 from transformers import CLIPImageProcessor, CLIPTextModel, CLIPTokenizer, CLIPVisionModelWithProjection
 
-from ...image_processor import PipelineImageInput, VaeImageProcessor
+from ...image_processor import PipelineImageInput
 from ...loaders import IPAdapterMixin, LoraLoaderMixin, TextualInversionLoaderMixin
 from ...models import AutoencoderKL, ImageProjection, UNet2DConditionModel, UNetMotionModel
 from ...models.lora import adjust_lora_scale_text_encoder
@@ -34,6 +33,7 @@ from ...schedulers import (
 )
 from ...utils import USE_PEFT_BACKEND, logging, scale_lora_layers, unscale_lora_layers
 from ...utils.torch_utils import randn_tensor
+from ...video_processor import VideoProcessor
 from ..free_init_utils import FreeInitMixin
 from ..pipeline_utils import DiffusionPipeline, StableDiffusionMixin
 from .pipeline_output import AnimateDiffPipelineOutput
@@ -52,14 +52,21 @@ EXAMPLE_DOC_STRING = """
         >>> from io import BytesIO
         >>> from PIL import Image
 
-        >>> adapter = MotionAdapter.from_pretrained("guoyww/animatediff-motion-adapter-v1-5-2", torch_dtype=torch.float16)
-        >>> pipe = AnimateDiffVideoToVideoPipeline.from_pretrained("SG161222/Realistic_Vision_V5.1_noVAE", motion_adapter=adapter).to("cuda")
-        >>> pipe.scheduler = DDIMScheduler(beta_schedule="linear", steps_offset=1, clip_sample=False, timespace_spacing="linspace")
+        >>> adapter = MotionAdapter.from_pretrained(
+        ...     "guoyww/animatediff-motion-adapter-v1-5-2", torch_dtype=torch.float16
+        ... )
+        >>> pipe = AnimateDiffVideoToVideoPipeline.from_pretrained(
+        ...     "SG161222/Realistic_Vision_V5.1_noVAE", motion_adapter=adapter
+        ... ).to("cuda")
+        >>> pipe.scheduler = DDIMScheduler(
+        ...     beta_schedule="linear", steps_offset=1, clip_sample=False, timespace_spacing="linspace"
+        ... )
+
 
         >>> def load_video(file_path: str):
         ...     images = []
-        ...
-        ...     if file_path.startswith(('http://', 'https://')):
+
+        ...     if file_path.startswith(("http://", "https://")):
         ...         # If the file_path is a URL
         ...         response = requests.get(file_path)
         ...         response.raise_for_status()
@@ -68,43 +75,26 @@ EXAMPLE_DOC_STRING = """
         ...     else:
         ...         # Assuming it's a local file path
         ...         vid = imageio.get_reader(file_path)
-        ...
+
         ...     for frame in vid:
         ...         pil_image = Image.fromarray(frame)
         ...         images.append(pil_image)
-        ...
+
         ...     return images
 
-        >>> video = load_video("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/animatediff-vid2vid-input-1.gif")
-        >>> output = pipe(video=video, prompt="panda playing a guitar, on a boat, in the ocean, high quality", strength=0.5)
+
+        >>> video = load_video(
+        ...     "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/animatediff-vid2vid-input-1.gif"
+        ... )
+        >>> output = pipe(
+        ...     video=video, prompt="panda playing a guitar, on a boat, in the ocean, high quality", strength=0.5
+        ... )
         >>> frames = output.frames[0]
         >>> export_to_gif(frames, "animation.gif")
         ```
 """
 
 
-# Copied from diffusers.pipelines.animatediff.pipeline_animatediff.tensor2vid
-def tensor2vid(video: torch.Tensor, processor, output_type="np"):
-    batch_size, channels, num_frames, height, width = video.shape
-    outputs = []
-    for batch_idx in range(batch_size):
-        batch_vid = video[batch_idx].permute(1, 0, 2, 3)
-        batch_output = processor.postprocess(batch_vid, output_type)
-
-        outputs.append(batch_output)
-
-    if output_type == "np":
-        outputs = np.stack(outputs)
-
-    elif output_type == "pt":
-        outputs = torch.stack(outputs)
-
-    elif not output_type == "pil":
-        raise ValueError(f"{output_type} does not exist. Please choose one of ['np', 'pt', 'pil']")
-
-    return outputs
-
-
 # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
 def retrieve_latents(
     encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
@@ -125,6 +115,7 @@ def retrieve_timesteps(
     num_inference_steps: Optional[int] = None,
     device: Optional[Union[str, torch.device]] = None,
     timesteps: Optional[List[int]] = None,
+    sigmas: Optional[List[float]] = None,
     **kwargs,
 ):
     """
@@ -135,19 +126,23 @@ def retrieve_timesteps(
         scheduler (`SchedulerMixin`):
             The scheduler to get timesteps from.
         num_inference_steps (`int`):
-            The number of diffusion steps used when generating samples with a pre-trained model. If used,
-            `timesteps` must be `None`.
+            The number of diffusion steps used when generating samples with a pre-trained model. If used, `timesteps`
+            must be `None`.
         device (`str` or `torch.device`, *optional*):
             The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
         timesteps (`List[int]`, *optional*):
-                Custom timesteps used to support arbitrary spacing between timesteps. If `None`, then the default
-                timestep spacing strategy of the scheduler is used. If `timesteps` is passed, `num_inference_steps`
-                must be `None`.
+            Custom timesteps used to override the timestep spacing strategy of the scheduler. If `timesteps` is passed,
+            `num_inference_steps` and `sigmas` must be `None`.
+        sigmas (`List[float]`, *optional*):
+            Custom sigmas used to override the timestep spacing strategy of the scheduler. If `sigmas` is passed,
+            `num_inference_steps` and `timesteps` must be `None`.
 
     Returns:
         `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
         second element is the number of inference steps.
     """
+    if timesteps is not None and sigmas is not None:
+        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
     if timesteps is not None:
         accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
         if not accepts_timesteps:
@@ -158,6 +153,16 @@ def retrieve_timesteps(
         scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
         timesteps = scheduler.timesteps
         num_inference_steps = len(timesteps)
+    elif sigmas is not None:
+        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accept_sigmas:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" sigmas schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
     else:
         scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
         timesteps = scheduler.timesteps
@@ -237,7 +242,7 @@ class AnimateDiffVideoToVideoPipeline(
             image_encoder=image_encoder,
         )
         self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
-        self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
+        self.video_processor = VideoProcessor(vae_scale_factor=self.vae_scale_factor)
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.encode_prompt with num_images_per_prompt -> num_videos_per_prompt
     def encode_prompt(
@@ -247,8 +252,8 @@ class AnimateDiffVideoToVideoPipeline(
         num_images_per_prompt,
         do_classifier_free_guidance,
         negative_prompt=None,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
         lora_scale: Optional[float] = None,
         clip_skip: Optional[int] = None,
     ):
@@ -268,10 +273,10 @@ class AnimateDiffVideoToVideoPipeline(
                 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*):
+            prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
                 provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
+            negative_prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
                 weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
                 argument.
@@ -623,16 +628,7 @@ class AnimateDiffVideoToVideoPipeline(
         generator,
         latents=None,
     ):
-        # video must be a list of list of images
-        # the outer list denotes having multiple videos as input, whereas inner list means the frames of the video
-        # as a list of images
-        if not isinstance(video[0], list):
-            video = [video]
         if latents is None:
-            video = torch.cat(
-                [self.image_processor.preprocess(vid, height=height, width=width).unsqueeze(0) for vid in video], dim=0
-            )
-            video = video.to(device=device, dtype=dtype)
             num_frames = video.shape[1]
         else:
             num_frames = latents.shape[2]
@@ -738,17 +734,18 @@ class AnimateDiffVideoToVideoPipeline(
         width: Optional[int] = None,
         num_inference_steps: int = 50,
         timesteps: Optional[List[int]] = None,
+        sigmas: Optional[List[float]] = None,
         guidance_scale: float = 7.5,
         strength: float = 0.8,
         negative_prompt: Optional[Union[str, List[str]]] = None,
         num_videos_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,
+        latents: Optional[torch.Tensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
         ip_adapter_image: Optional[PipelineImageInput] = None,
-        ip_adapter_image_embeds: Optional[List[torch.FloatTensor]] = None,
+        ip_adapter_image_embeds: Optional[List[torch.Tensor]] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
         cross_attention_kwargs: Optional[Dict[str, Any]] = None,
@@ -771,6 +768,14 @@ class AnimateDiffVideoToVideoPipeline(
             num_inference_steps (`int`, *optional*, defaults to 50):
                 The number of denoising steps. More denoising steps usually lead to a higher quality videos at the
                 expense of slower inference.
+            timesteps (`List[int]`, *optional*):
+                Custom timesteps to use for the denoising process with schedulers which support a `timesteps` argument
+                in their `set_timesteps` method. If not defined, the default behavior when `num_inference_steps` is
+                passed will be used. Must be in descending order.
+            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.
             strength (`float`, *optional*, defaults to 0.8):
                 Higher strength leads to more differences between original video and generated video.
             guidance_scale (`float`, *optional*, defaults to 7.5):
@@ -785,30 +790,28 @@ class AnimateDiffVideoToVideoPipeline(
             generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
                 A [`torch.Generator`](https://pytorch.org/docs/stable/generated/torch.Generator.html) to make
                 generation deterministic.
-            latents (`torch.FloatTensor`, *optional*):
+            latents (`torch.Tensor`, *optional*):
                 Pre-generated noisy latents sampled from a Gaussian distribution, to be used as inputs for video
                 generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
                 tensor is generated by sampling using the supplied random `generator`. Latents should be of shape
                 `(batch_size, num_channel, num_frames, height, width)`.
-            prompt_embeds (`torch.FloatTensor`, *optional*):
+            prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated text embeddings. Can be used to easily tweak text inputs (prompt weighting). If not
                 provided, text embeddings are generated from the `prompt` input argument.
-            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
+            negative_prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated negative text embeddings. Can be used to easily tweak text inputs (prompt weighting). If
                 not provided, `negative_prompt_embeds` are generated from the `negative_prompt` input argument.
             ip_adapter_image: (`PipelineImageInput`, *optional*):
                 Optional image input to work with IP Adapters.
-            ip_adapter_image_embeds (`List[torch.FloatTensor]`, *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)`. It should contain the negative image embedding
-                if `do_classifier_free_guidance` is set to `True`.
-                If not provided, embeddings are computed from the `ip_adapter_image` input argument.
+            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)`. It should
+                contain the negative image embedding if `do_classifier_free_guidance` is set to `True`. If not
+                provided, embeddings are computed from the `ip_adapter_image` input argument.
             output_type (`str`, *optional*, defaults to `"pil"`):
-                The output format of the generated video. Choose between `torch.FloatTensor`, `PIL.Image` or
-                `np.array`.
+                The output format of the generated video. Choose between `torch.Tensor`, `PIL.Image` or `np.array`.
             return_dict (`bool`, *optional*, defaults to `True`):
-                Whether or not to return a [`AnimateDiffPipelineOutput`] instead
-                of a plain tuple.
+                Whether or not to return a [`AnimateDiffPipelineOutput`] instead of a plain tuple.
             cross_attention_kwargs (`dict`, *optional*):
                 A kwargs dictionary that if specified is passed along to the [`AttentionProcessor`] as defined in
                 [`self.processor`](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
@@ -823,7 +826,7 @@ class AnimateDiffVideoToVideoPipeline(
             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 pipeine class.
+                `._callback_tensor_inputs` attribute of your pipeline class.
 
         Examples:
 
@@ -901,11 +904,18 @@ class AnimateDiffVideoToVideoPipeline(
             )
 
         # 4. Prepare timesteps
-        timesteps, num_inference_steps = retrieve_timesteps(self.scheduler, num_inference_steps, device, timesteps)
+        timesteps, num_inference_steps = retrieve_timesteps(
+            self.scheduler, num_inference_steps, device, timesteps, sigmas
+        )
         timesteps, num_inference_steps = self.get_timesteps(num_inference_steps, timesteps, strength, device)
         latent_timestep = timesteps[:1].repeat(batch_size * num_videos_per_prompt)
 
         # 5. Prepare latent variables
+        if latents is None:
+            video = self.video_processor.preprocess_video(video, height=height, width=width)
+            # Move the number of frames before the number of channels.
+            video = video.permute(0, 2, 1, 3, 4)
+            video = video.to(device=device, dtype=prompt_embeds.dtype)
         num_channels_latents = self.unet.config.in_channels
         latents = self.prepare_latents(
             video=video,
@@ -944,7 +954,7 @@ class AnimateDiffVideoToVideoPipeline(
             num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
 
             # 8. Denoising loop
-            with self.progress_bar(total=num_inference_steps) as progress_bar:
+            with self.progress_bar(total=self._num_timesteps) as progress_bar:
                 for i, t in enumerate(timesteps):
                     # expand the latents if we are doing classifier free guidance
                     latent_model_input = torch.cat([latents] * 2) if self.do_classifier_free_guidance else latents
@@ -986,7 +996,7 @@ class AnimateDiffVideoToVideoPipeline(
             video = latents
         else:
             video_tensor = self.decode_latents(latents)
-            video = tensor2vid(video_tensor, self.image_processor, output_type=output_type)
+            video = self.video_processor.postprocess_video(video=video_tensor, output_type=output_type)
 
         # 10. Offload all models
         self.maybe_free_model_hooks()

diffusers/pipelines/animatediff/pipeline_output.py

@@ -13,9 +13,10 @@ class AnimateDiffPipelineOutput(BaseOutput):
     r"""
      Output class for AnimateDiff pipelines.
 
-     Args:
+    Args:
          frames (`torch.Tensor`, `np.ndarray`, or List[List[PIL.Image.Image]]):
-             List of video outputs - It can be a nested list of length `batch_size,` with each sub-list containing denoised
+             List of video outputs - It can be a nested list of length `batch_size,` with each sub-list containing
+             denoised
      PIL image sequences of length `num_frames.` It can also be a NumPy array or Torch tensor of shape
     `(batch_size, num_frames, channels, height, width)`
     """

diffusers/pipelines/audioldm/pipeline_audioldm.py

@@ -103,8 +103,8 @@ class AudioLDMPipeline(DiffusionPipeline, StableDiffusionMixin):
         num_waveforms_per_prompt,
         do_classifier_free_guidance,
         negative_prompt=None,
-        prompt_embeds: Optional[torch.FloatTensor] = None,
-        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
     ):
         r"""
         Encodes the prompt into text encoder hidden states.
@@ -122,10 +122,10 @@ class AudioLDMPipeline(DiffusionPipeline, StableDiffusionMixin):
                 The prompt or prompts not to guide the audio 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*):
+            prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
                 provided, text embeddings will be generated from `prompt` input argument.
-            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
+            negative_prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
                 weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
                 argument.
@@ -330,8 +330,8 @@ class AudioLDMPipeline(DiffusionPipeline, StableDiffusionMixin):
         shape = (
             batch_size,
             num_channels_latents,
-            height // self.vae_scale_factor,
-            self.vocoder.config.model_in_dim // self.vae_scale_factor,
+            int(height) // self.vae_scale_factor,
+            int(self.vocoder.config.model_in_dim) // self.vae_scale_factor,
         )
         if isinstance(generator, list) and len(generator) != batch_size:
             raise ValueError(
@@ -360,11 +360,11 @@ class AudioLDMPipeline(DiffusionPipeline, StableDiffusionMixin):
         num_waveforms_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,
+        latents: Optional[torch.Tensor] = None,
+        prompt_embeds: Optional[torch.Tensor] = None,
+        negative_prompt_embeds: Optional[torch.Tensor] = None,
         return_dict: bool = True,
-        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback: Optional[Callable[[int, int, torch.Tensor], None]] = None,
         callback_steps: Optional[int] = 1,
         cross_attention_kwargs: Optional[Dict[str, Any]] = None,
         output_type: Optional[str] = "np",
@@ -394,21 +394,21 @@ class AudioLDMPipeline(DiffusionPipeline, StableDiffusionMixin):
             generator (`torch.Generator` or `List[torch.Generator]`, *optional*):
                 A [`torch.Generator`](https://pytorch.org/docs/stable/generated/torch.Generator.html) to make
                 generation deterministic.
-            latents (`torch.FloatTensor`, *optional*):
+            latents (`torch.Tensor`, *optional*):
                 Pre-generated noisy latents sampled from a Gaussian distribution, to be used as inputs for image
                 generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
                 tensor is generated by sampling using the supplied random `generator`.
-            prompt_embeds (`torch.FloatTensor`, *optional*):
+            prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated text embeddings. Can be used to easily tweak text inputs (prompt weighting). If not
                 provided, text embeddings are generated from the `prompt` input argument.
-            negative_prompt_embeds (`torch.FloatTensor`, *optional*):
+            negative_prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated negative text embeddings. Can be used to easily tweak text inputs (prompt weighting). If
                 not provided, `negative_prompt_embeds` are generated from the `negative_prompt` input argument.
             return_dict (`bool`, *optional*, defaults to `True`):
                 Whether or not to return a [`~pipelines.AudioPipelineOutput`] instead of a plain tuple.
             callback (`Callable`, *optional*):
                 A function that calls every `callback_steps` steps during inference. The function is called with the
-                following arguments: `callback(step: int, timestep: int, latents: torch.FloatTensor)`.
+                following arguments: `callback(step: int, timestep: int, latents: torch.Tensor)`.
             callback_steps (`int`, *optional*, defaults to 1):
                 The frequency at which the `callback` function is called. If not specified, the callback is called at
                 every step.

diffusers/pipelines/audioldm2/modeling_audioldm2.py

@@ -64,7 +64,7 @@ class AudioLDM2ProjectionModelOutput(BaseOutput):
     """
     Args:
     Class for AudioLDM2 projection layer's outputs.
-        hidden_states (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`):
+        hidden_states (`torch.Tensor` of shape `(batch_size, sequence_length, hidden_size)`):
             Sequence of hidden-states obtained by linearly projecting the hidden-states for each of the text
              encoders and subsequently concatenating them together.
         attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
@@ -75,7 +75,7 @@ class AudioLDM2ProjectionModelOutput(BaseOutput):
             - 0 for tokens that are **masked**.
     """
 
-    hidden_states: torch.FloatTensor
+    hidden_states: torch.Tensor
     attention_mask: Optional[torch.LongTensor] = None
 
 
@@ -95,7 +95,14 @@ class AudioLDM2ProjectionModel(ModelMixin, ConfigMixin):
     """
 
     @register_to_config
-    def __init__(self, text_encoder_dim, text_encoder_1_dim, langauge_model_dim):
+    def __init__(
+        self,
+        text_encoder_dim,
+        text_encoder_1_dim,
+        langauge_model_dim,
+        use_learned_position_embedding=None,
+        max_seq_length=None,
+    ):
         super().__init__()
         # additional projection layers for each text encoder
         self.projection = nn.Linear(text_encoder_dim, langauge_model_dim)
@@ -108,10 +115,18 @@ class AudioLDM2ProjectionModel(ModelMixin, ConfigMixin):
         self.sos_embed_1 = nn.Parameter(torch.ones(langauge_model_dim))
         self.eos_embed_1 = nn.Parameter(torch.ones(langauge_model_dim))
 
+        self.use_learned_position_embedding = use_learned_position_embedding
+
+        # learable positional embedding for vits encoder
+        if self.use_learned_position_embedding is not None:
+            self.learnable_positional_embedding = torch.nn.Parameter(
+                torch.zeros((1, text_encoder_1_dim, max_seq_length))
+            )
+
     def forward(
         self,
-        hidden_states: Optional[torch.FloatTensor] = None,
-        hidden_states_1: Optional[torch.FloatTensor] = None,
+        hidden_states: Optional[torch.Tensor] = None,
+        hidden_states_1: Optional[torch.Tensor] = None,
         attention_mask: Optional[torch.LongTensor] = None,
         attention_mask_1: Optional[torch.LongTensor] = None,
     ):
@@ -120,6 +135,10 @@ class AudioLDM2ProjectionModel(ModelMixin, ConfigMixin):
             hidden_states, attention_mask, sos_token=self.sos_embed, eos_token=self.eos_embed
         )
 
+        # Add positional embedding for Vits hidden state
+        if self.use_learned_position_embedding is not None:
+            hidden_states_1 = (hidden_states_1.permute(0, 2, 1) + self.learnable_positional_embedding).permute(0, 2, 1)
+
         hidden_states_1 = self.projection_1(hidden_states_1)
         hidden_states_1, attention_mask_1 = add_special_tokens(
             hidden_states_1, attention_mask_1, sos_token=self.sos_embed_1, eos_token=self.eos_embed_1
@@ -661,7 +680,7 @@ class AudioLDM2UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoad
 
     def forward(
         self,
-        sample: torch.FloatTensor,
+        sample: torch.Tensor,
         timestep: Union[torch.Tensor, float, int],
         encoder_hidden_states: torch.Tensor,
         class_labels: Optional[torch.Tensor] = None,
@@ -677,10 +696,10 @@ class AudioLDM2UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoad
         The [`AudioLDM2UNet2DConditionModel`] forward method.
 
         Args:
-            sample (`torch.FloatTensor`):
+            sample (`torch.Tensor`):
                 The noisy input tensor with the following shape `(batch, channel, height, width)`.
-            timestep (`torch.FloatTensor` or `float` or `int`): The number of timesteps to denoise an input.
-            encoder_hidden_states (`torch.FloatTensor`):
+            timestep (`torch.Tensor` or `float` or `int`): The number of timesteps to denoise an input.
+            encoder_hidden_states (`torch.Tensor`):
                 The encoder hidden states with shape `(batch, sequence_length, feature_dim)`.
             encoder_attention_mask (`torch.Tensor`):
                 A cross-attention mask of shape `(batch, sequence_length)` is applied to `encoder_hidden_states`. If
@@ -691,7 +710,7 @@ class AudioLDM2UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoad
                 tuple.
             cross_attention_kwargs (`dict`, *optional*):
                 A kwargs dictionary that if specified is passed along to the [`AttnProcessor`].
-            encoder_hidden_states_1 (`torch.FloatTensor`, *optional*):
+            encoder_hidden_states_1 (`torch.Tensor`, *optional*):
                 A second set of encoder hidden states with shape `(batch, sequence_length_2, feature_dim_2)`. Can be
                 used to condition the model on a different set of embeddings to `encoder_hidden_states`.
             encoder_attention_mask_1 (`torch.Tensor`, *optional*):
@@ -701,8 +720,8 @@ class AudioLDM2UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoad
 
         Returns:
             [`~models.unets.unet_2d_condition.UNet2DConditionOutput`] or `tuple`:
-                If `return_dict` is True, an [`~models.unets.unet_2d_condition.UNet2DConditionOutput`] is returned, otherwise
-                a `tuple` is returned where the first element is the sample tensor.
+                If `return_dict` is True, an [`~models.unets.unet_2d_condition.UNet2DConditionOutput`] is returned,
+                otherwise a `tuple` is returned where the first element is the sample tensor.
         """
         # By default samples have to be AT least a multiple of the overall upsampling factor.
         # The overall upsampling factor is equal to 2 ** (# num of upsampling layers).
@@ -1072,14 +1091,14 @@ class CrossAttnDownBlock2D(nn.Module):
 
     def forward(
         self,
-        hidden_states: torch.FloatTensor,
-        temb: Optional[torch.FloatTensor] = None,
-        encoder_hidden_states: Optional[torch.FloatTensor] = None,
-        attention_mask: Optional[torch.FloatTensor] = None,
+        hidden_states: torch.Tensor,
+        temb: Optional[torch.Tensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
         cross_attention_kwargs: Optional[Dict[str, Any]] = None,
-        encoder_attention_mask: Optional[torch.FloatTensor] = None,
-        encoder_hidden_states_1: Optional[torch.FloatTensor] = None,
-        encoder_attention_mask_1: Optional[torch.FloatTensor] = None,
+        encoder_attention_mask: Optional[torch.Tensor] = None,
+        encoder_hidden_states_1: Optional[torch.Tensor] = None,
+        encoder_attention_mask_1: Optional[torch.Tensor] = None,
     ):
         output_states = ()
         num_layers = len(self.resnets)
@@ -1251,15 +1270,15 @@ class UNetMidBlock2DCrossAttn(nn.Module):
 
     def forward(
         self,
-        hidden_states: torch.FloatTensor,
-        temb: Optional[torch.FloatTensor] = None,
-        encoder_hidden_states: Optional[torch.FloatTensor] = None,
-        attention_mask: Optional[torch.FloatTensor] = None,
+        hidden_states: torch.Tensor,
+        temb: Optional[torch.Tensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
         cross_attention_kwargs: Optional[Dict[str, Any]] = None,
-        encoder_attention_mask: Optional[torch.FloatTensor] = None,
-        encoder_hidden_states_1: Optional[torch.FloatTensor] = None,
-        encoder_attention_mask_1: Optional[torch.FloatTensor] = None,
-    ) -> torch.FloatTensor:
+        encoder_attention_mask: Optional[torch.Tensor] = None,
+        encoder_hidden_states_1: Optional[torch.Tensor] = None,
+        encoder_attention_mask_1: Optional[torch.Tensor] = None,
+    ) -> torch.Tensor:
         hidden_states = self.resnets[0](hidden_states, temb)
         num_attention_per_layer = len(self.attentions) // (len(self.resnets) - 1)
 
@@ -1418,16 +1437,16 @@ class CrossAttnUpBlock2D(nn.Module):
 
     def forward(
         self,
-        hidden_states: torch.FloatTensor,
-        res_hidden_states_tuple: Tuple[torch.FloatTensor, ...],
-        temb: Optional[torch.FloatTensor] = None,
-        encoder_hidden_states: Optional[torch.FloatTensor] = None,
+        hidden_states: torch.Tensor,
+        res_hidden_states_tuple: Tuple[torch.Tensor, ...],
+        temb: Optional[torch.Tensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
         cross_attention_kwargs: Optional[Dict[str, Any]] = None,
         upsample_size: Optional[int] = None,
-        attention_mask: Optional[torch.FloatTensor] = None,
-        encoder_attention_mask: Optional[torch.FloatTensor] = None,
-        encoder_hidden_states_1: Optional[torch.FloatTensor] = None,
-        encoder_attention_mask_1: Optional[torch.FloatTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        encoder_attention_mask: Optional[torch.Tensor] = None,
+        encoder_hidden_states_1: Optional[torch.Tensor] = None,
+        encoder_attention_mask_1: Optional[torch.Tensor] = None,
     ):
         num_layers = len(self.resnets)
         num_attention_per_layer = len(self.attentions) // num_layers