HunyuanVideo

HunyuanVideo by Tencent.

Recent advancements in video generation have significantly impacted daily life for both individuals and industries. However, the leading video generation models remain closed-source, resulting in a notable performance gap between industry capabilities and those available to the public. In this report, we introduce HunyuanVideo, an innovative open-source video foundation model that demonstrates performance in video generation comparable to, or even surpassing, that of leading closed-source models. HunyuanVideo encompasses a comprehensive framework that integrates several key elements, including data curation, advanced architectural design, progressive model scaling and training, and an efficient infrastructure tailored for large-scale model training and inference. As a result, we successfully trained a video generative model with over 13 billion parameters, making it the largest among all open-source models. We conducted extensive experiments and implemented a series of targeted designs to ensure high visual quality, motion dynamics, text-video alignment, and advanced filming techniques. According to evaluations by professionals, HunyuanVideo outperforms previous state-of-the-art models, including Runway Gen-3, Luma 1.6, and three top-performing Chinese video generative models. By releasing the code for the foundation model and its applications, we aim to bridge the gap between closed-source and open-source communities. This initiative will empower individuals within the community to experiment with their ideas, fostering a more dynamic and vibrant video generation ecosystem. The code is publicly available at this https URL.

Tip

Make sure to check out the Schedulers guide to learn how to explore the tradeoff between scheduler speed and quality.

Recommendations for inference: - Both text encoders should be in ms.float16. - Transformer should be in ms.bfloat16. - VAE should be in ms.float16. - num_frames should be of the form 4 * k + 1, for example 49 or 129. - For smaller resolution videos, try lower values of shift (between 2.0 to 5.0) in the Scheduler. For larger resolution images, try higher values (between 7.0 and 12.0). The default value is 7.0 for HunyuanVideo. - For more information about supported resolutions and other details, please refer to the original repository here.

mindone.diffusers.HunyuanVideoPipeline

Bases: DiffusionPipeline, HunyuanVideoLoraLoaderMixin

Pipeline for text-to-video generation using HunyuanVideo.

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

PARAMETER	DESCRIPTION
text_encoder	Llava Llama3-8B. TYPE: [`LlamaModel`]
tokenizer	Tokenizer from Llava Llama3-8B. TYPE: `LlamaTokenizer`
transformer	Conditional Transformer to denoise the encoded image latents. TYPE: [`HunyuanVideoTransformer3DModel`]
scheduler	A scheduler to be used in combination with transformer to denoise the encoded image latents. TYPE: [`FlowMatchEulerDiscreteScheduler`]
vae	Variational Auto-Encoder (VAE) Model to encode and decode videos to and from latent representations. TYPE: [`AutoencoderKLHunyuanVideo`]
text_encoder_2	CLIP, specifically the clip-vit-large-patch14 variant. TYPE: [`CLIPTextModel`]
tokenizer_2	Tokenizer of class CLIPTokenizer. TYPE: `CLIPTokenizer`

Source code in mindone/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video.py

class HunyuanVideoPipeline(DiffusionPipeline, HunyuanVideoLoraLoaderMixin):
    r"""
    Pipeline for text-to-video generation using HunyuanVideo.

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

    Args:
        text_encoder ([`LlamaModel`]):
            [Llava Llama3-8B](https://huggingface.co/xtuner/llava-llama-3-8b-v1_1-transformers).
        tokenizer (`LlamaTokenizer`):
            Tokenizer from [Llava Llama3-8B](https://huggingface.co/xtuner/llava-llama-3-8b-v1_1-transformers).
        transformer ([`HunyuanVideoTransformer3DModel`]):
            Conditional Transformer to denoise the encoded image latents.
        scheduler ([`FlowMatchEulerDiscreteScheduler`]):
            A scheduler to be used in combination with `transformer` to denoise the encoded image latents.
        vae ([`AutoencoderKLHunyuanVideo`]):
            Variational Auto-Encoder (VAE) Model to encode and decode videos to and from latent representations.
        text_encoder_2 ([`CLIPTextModel`]):
            [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_2 (`CLIPTokenizer`):
            Tokenizer of class
            [CLIPTokenizer](https://huggingface.co/docs/transformers/en/model_doc/clip#transformers.CLIPTokenizer).
    """

    model_cpu_offload_seq = "text_encoder->text_encoder_2->transformer->vae"
    _callback_tensor_inputs = ["latents", "prompt_embeds"]

    def __init__(
        self,
        text_encoder: LlamaModel,
        tokenizer: LlamaTokenizerFast,
        transformer: HunyuanVideoTransformer3DModel,
        vae: AutoencoderKLHunyuanVideo,
        scheduler: FlowMatchEulerDiscreteScheduler,
        text_encoder_2: CLIPTextModel,
        tokenizer_2: CLIPTokenizer,
    ):
        super().__init__()

        self.register_modules(
            vae=vae,
            text_encoder=text_encoder,
            tokenizer=tokenizer,
            transformer=transformer,
            scheduler=scheduler,
            text_encoder_2=text_encoder_2,
            tokenizer_2=tokenizer_2,
        )

        self.vae_scale_factor_temporal = (
            self.vae.temporal_compression_ratio if hasattr(self, "vae") and self.vae is not None else 4
        )
        self.vae_scale_factor_spatial = (
            self.vae.spatial_compression_ratio if hasattr(self, "vae") and self.vae is not None else 8
        )
        self.video_processor = VideoProcessor(vae_scale_factor=self.vae_scale_factor_spatial)

    def _get_llama_prompt_embeds(
        self,
        prompt: Union[str, List[str]],
        prompt_template: Dict[str, Any],
        num_videos_per_prompt: int = 1,
        dtype: Optional[ms.Type] = None,
        max_sequence_length: int = 256,
        num_hidden_layers_to_skip: int = 2,
    ) -> Tuple[ms.Tensor, ms.Tensor]:
        dtype = dtype or self.text_encoder.dtype

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

        prompt = [prompt_template["template"].format(p) for p in prompt]

        crop_start = prompt_template.get("crop_start", None)
        if crop_start is None:
            prompt_template_input = self.tokenizer(
                prompt_template["template"],
                padding="max_length",
                return_tensors="np",
                return_length=False,
                return_overflowing_tokens=False,
                return_attention_mask=False,
            )
            crop_start = prompt_template_input["input_ids"].shape[-1]
            # Remove <|eot_id|> token and placeholder {}
            crop_start -= 2

        max_sequence_length += crop_start
        text_inputs = self.tokenizer(
            prompt,
            max_length=max_sequence_length,
            padding="max_length",
            truncation=True,
            return_tensors="np",
            return_length=False,
            return_overflowing_tokens=False,
            return_attention_mask=True,
        )
        text_input_ids = ms.Tensor(text_inputs.input_ids)
        prompt_attention_mask = ms.Tensor(text_inputs.attention_mask)

        prompt_embeds = self.text_encoder(
            input_ids=text_input_ids,
            attention_mask=prompt_attention_mask,
            output_hidden_states=True,
        )[2][-(num_hidden_layers_to_skip + 1)]
        prompt_embeds = prompt_embeds.to(dtype=dtype)

        if crop_start is not None and crop_start > 0:
            prompt_embeds = prompt_embeds[:, crop_start:]
            prompt_attention_mask = prompt_attention_mask[:, crop_start:]

        # duplicate text embeddings for each generation per prompt, using mps friendly method
        _, seq_len, _ = prompt_embeds.shape
        prompt_embeds = prompt_embeds.tile((1, num_videos_per_prompt, 1))
        prompt_embeds = prompt_embeds.view(batch_size * num_videos_per_prompt, seq_len, -1)
        prompt_attention_mask = prompt_attention_mask.tile((1, num_videos_per_prompt))
        prompt_attention_mask = prompt_attention_mask.view(batch_size * num_videos_per_prompt, seq_len)

        return prompt_embeds, prompt_attention_mask

    def _get_clip_prompt_embeds(
        self,
        prompt: Union[str, List[str]],
        num_videos_per_prompt: int = 1,
        dtype: Optional[ms.Type] = None,
        max_sequence_length: int = 77,
    ) -> ms.Tensor:
        dtype = dtype or self.text_encoder_2.dtype

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

        text_inputs = self.tokenizer_2(
            prompt,
            padding="max_length",
            max_length=max_sequence_length,
            truncation=True,
            return_tensors="np",
        )

        text_input_ids = text_inputs.input_ids
        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="np").input_ids
        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not np.array_equal(
            text_input_ids, untruncated_ids
        ):
            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, max_sequence_length - 1 : -1])
            logger.warning(
                "The following part of your input was truncated because CLIP can only handle sequences up to"
                f" {max_sequence_length} tokens: {removed_text}"
            )

        prompt_embeds = self.text_encoder_2(ms.Tensor(text_input_ids), output_hidden_states=False)[1]

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

        return prompt_embeds

    def encode_prompt(
        self,
        prompt: Union[str, List[str]],
        prompt_2: Union[str, List[str]] = None,
        prompt_template: Dict[str, Any] = DEFAULT_PROMPT_TEMPLATE,
        num_videos_per_prompt: int = 1,
        prompt_embeds: Optional[ms.Tensor] = None,
        pooled_prompt_embeds: Optional[ms.Tensor] = None,
        prompt_attention_mask: Optional[ms.Tensor] = None,
        dtype: Optional[ms.Type] = None,
        max_sequence_length: int = 256,
    ):
        if prompt_embeds is None:
            prompt_embeds, prompt_attention_mask = self._get_llama_prompt_embeds(
                prompt,
                prompt_template,
                num_videos_per_prompt,
                dtype=dtype,
                max_sequence_length=max_sequence_length,
            )

        if pooled_prompt_embeds is None:
            if prompt_2 is None and pooled_prompt_embeds is None:
                prompt_2 = prompt
            pooled_prompt_embeds = self._get_clip_prompt_embeds(
                prompt,
                num_videos_per_prompt,
                dtype=dtype,
                max_sequence_length=77,
            )

        return prompt_embeds, pooled_prompt_embeds, prompt_attention_mask

    def check_inputs(
        self,
        prompt,
        prompt_2,
        height,
        width,
        prompt_embeds=None,
        callback_on_step_end_tensor_inputs=None,
        prompt_template=None,
    ):
        if height % 16 != 0 or width % 16 != 0:
            raise ValueError(f"`height` and `width` have to be divisible by 8 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
        ):
            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]}"  # noqa: E501
            )

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

        if prompt_template is not None:
            if not isinstance(prompt_template, dict):
                raise ValueError(f"`prompt_template` has to be of type `dict` but is {type(prompt_template)}")
            if "template" not in prompt_template:
                raise ValueError(
                    f"`prompt_template` has to contain a key `template` but only found {prompt_template.keys()}"
                )

    def prepare_latents(
        self,
        batch_size: int,
        num_channels_latents: 32,
        height: int = 720,
        width: int = 1280,
        num_frames: int = 129,
        dtype: Optional[ms.Type] = None,
        generator: Optional[Union[np.random.Generator, List[np.random.Generator]]] = None,
        latents: Optional[ms.Tensor] = None,
    ) -> ms.Tensor:
        if latents is not None:
            return latents.to(dtype=dtype)

        shape = (
            batch_size,
            num_channels_latents,
            num_frames,
            int(height) // self.vae_scale_factor_spatial,
            int(width) // self.vae_scale_factor_spatial,
        )
        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."
            )

        latents = randn_tensor(shape, generator=generator, dtype=dtype)
        return latents

    def enable_vae_slicing(self):
        r"""
        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
        """
        self.vae.enable_slicing()

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

    def enable_vae_tiling(self):
        r"""
        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
        processing larger images.
        """
        self.vae.enable_tiling()

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

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

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

    @property
    def attention_kwargs(self):
        return self._attention_kwargs

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

    def __call__(
        self,
        prompt: Union[str, List[str]] = None,
        prompt_2: Union[str, List[str]] = None,
        height: int = 720,
        width: int = 1280,
        num_frames: int = 129,
        num_inference_steps: int = 50,
        sigmas: List[float] = None,
        guidance_scale: float = 6.0,
        num_videos_per_prompt: Optional[int] = 1,
        generator: Optional[Union[np.random.Generator, List[np.random.Generator]]] = None,
        latents: Optional[ms.Tensor] = None,
        prompt_embeds: Optional[ms.Tensor] = None,
        pooled_prompt_embeds: Optional[ms.Tensor] = None,
        prompt_attention_mask: Optional[ms.Tensor] = None,
        output_type: Optional[str] = "pil",
        return_dict: bool = False,
        attention_kwargs: Optional[Dict[str, Any]] = None,
        callback_on_step_end: Optional[
            Union[Callable[[int, int, Dict], None], PipelineCallback, MultiPipelineCallbacks]
        ] = None,
        callback_on_step_end_tensor_inputs: List[str] = ["latents"],
        prompt_template: Dict[str, Any] = DEFAULT_PROMPT_TEMPLATE,
        max_sequence_length: int = 256,
    ):
        r"""
        The call function to 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.
            height (`int`, defaults to `720`):
                The height in pixels of the generated image.
            width (`int`, defaults to `1280`):
                The width in pixels of the generated image.
            num_frames (`int`, defaults to `129`):
                The number of frames in the generated video.
            num_inference_steps (`int`, 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`, defaults to `6.0`):
                Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
                `guidance_scale` is defined as `w` of equation 2. of [Imagen
                Paper](https://arxiv.org/pdf/2205.11487.pdf). Guidance scale is enabled by setting `guidance_scale >
                1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`,
                usually at the expense of lower image quality. Note that the only available HunyuanVideo model is
                CFG-distilled, which means that traditional guidance between unconditional and conditional latent is
                not applied.
            num_videos_per_prompt (`int`, *optional*, defaults to 1):
                The number of images to generate per prompt.
            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.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.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.
            output_type (`str`, *optional*, defaults to `"pil"`):
                The output format of the generated image. Choose between `PIL.Image` or `np.array`.
            return_dict (`bool`, *optional*, defaults to `False`):
                Whether or not to return a [`HunyuanVideoPipelineOutput`] instead of a plain tuple.
            attention_kwargs (`dict`, *optional*):
                A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
                `self.processor` in
                [diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
            clip_skip (`int`, *optional*):
                Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
                the output of the pre-final layer will be used for computing the prompt embeddings.
            callback_on_step_end (`Callable`, `PipelineCallback`, `MultiPipelineCallbacks`, *optional*):
                A function or a subclass of `PipelineCallback` or `MultiPipelineCallbacks` that is called at the end of
                each denoising step during the inference. 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.

        Examples:

        Returns:
            [`~HunyuanVideoPipelineOutput`] or `tuple`:
                If `return_dict` is `True`, [`HunyuanVideoPipelineOutput`] is returned, otherwise a `tuple` is returned
                where the first element is a list with the generated images and the second element is a list of `bool`s
                indicating whether the corresponding generated image contains "not-safe-for-work" (nsfw) content.
        """

        if isinstance(callback_on_step_end, (PipelineCallback, MultiPipelineCallbacks)):
            callback_on_step_end_tensor_inputs = callback_on_step_end.tensor_inputs

        # 1. Check inputs. Raise error if not correct
        self.check_inputs(
            prompt,
            prompt_2,
            height,
            width,
            prompt_embeds,
            callback_on_step_end_tensor_inputs,
            prompt_template,
        )

        self._guidance_scale = guidance_scale
        self._attention_kwargs = attention_kwargs
        self._interrupt = False

        # 2. Define call parameters
        if prompt is not None and isinstance(prompt, str):
            batch_size = 1
        elif prompt is not None and isinstance(prompt, list):
            batch_size = len(prompt)
        else:
            batch_size = prompt_embeds.shape[0]

        # 3. Encode input prompt
        prompt_embeds, pooled_prompt_embeds, prompt_attention_mask = self.encode_prompt(
            prompt=prompt,
            prompt_2=prompt_2,
            prompt_template=prompt_template,
            num_videos_per_prompt=num_videos_per_prompt,
            prompt_embeds=prompt_embeds,
            pooled_prompt_embeds=pooled_prompt_embeds,
            prompt_attention_mask=prompt_attention_mask,
            max_sequence_length=max_sequence_length,
        )

        transformer_dtype = self.transformer.dtype
        prompt_embeds = prompt_embeds.to(transformer_dtype)
        prompt_attention_mask = prompt_attention_mask.to(transformer_dtype)
        if pooled_prompt_embeds is not None:
            pooled_prompt_embeds = pooled_prompt_embeds.to(transformer_dtype)

        # 4. Prepare timesteps
        sigmas = np.linspace(1.0, 0.0, num_inference_steps + 1)[:-1] if sigmas is None else sigmas
        timesteps, num_inference_steps = retrieve_timesteps(
            self.scheduler,
            num_inference_steps,
            sigmas=sigmas,
        )

        # 5. Prepare latent variables
        num_channels_latents = self.transformer.config.in_channels
        num_latent_frames = (num_frames - 1) // self.vae_scale_factor_temporal + 1
        latents = self.prepare_latents(
            batch_size * num_videos_per_prompt,
            num_channels_latents,
            height,
            width,
            num_latent_frames,
            ms.float32,
            generator,
            latents,
        )

        # 6. Prepare guidance condition
        guidance = ms.Tensor([guidance_scale] * latents.shape[0], dtype=transformer_dtype) * 1000.0

        # 7. Denoising loop
        num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
        self._num_timesteps = len(timesteps)

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

                latent_model_input = latents.to(transformer_dtype)
                # broadcast to batch dimension in a way that's compatible with ONNX/Core ML
                timestep = t.broadcast_to((latents.shape[0],)).to(latents.dtype)

                noise_pred = self.transformer(
                    hidden_states=latent_model_input,
                    timestep=timestep,
                    encoder_hidden_states=prompt_embeds,
                    encoder_attention_mask=prompt_attention_mask,
                    pooled_projections=pooled_prompt_embeds,
                    guidance=guidance,
                    attention_kwargs=attention_kwargs,
                    return_dict=False,
                )[0]

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

                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 not output_type == "latent":
            latents = latents.to(self.vae.dtype) / self.vae.config.scaling_factor
            video = self.vae.decode(latents, return_dict=False)[0]
            video = self.video_processor.postprocess_video(video, output_type=output_type)
        else:
            video = latents

        if not return_dict:
            return (video,)

        return HunyuanVideoPipelineOutput(frames=video)

mindone.diffusers.HunyuanVideoPipeline.__call__(prompt=None, prompt_2=None, height=720, width=1280, num_frames=129, num_inference_steps=50, sigmas=None, guidance_scale=6.0, num_videos_per_prompt=1, generator=None, latents=None, prompt_embeds=None, pooled_prompt_embeds=None, prompt_attention_mask=None, output_type='pil', return_dict=False, attention_kwargs=None, callback_on_step_end=None, callback_on_step_end_tensor_inputs=['latents'], prompt_template=DEFAULT_PROMPT_TEMPLATE, max_sequence_length=256)

The call function to the pipeline for generation.

PARAMETER	DESCRIPTION
prompt	The prompt or prompts to guide the image generation. If not defined, one has to pass prompt_embeds. instead. TYPE: `str` or `List[str]`, *optional* DEFAULT: None
prompt_2	The prompt or prompts to be sent to tokenizer_2 and text_encoder_2. If not defined, prompt is will be used instead. TYPE: `str` or `List[str]`, *optional* DEFAULT: None
height	The height in pixels of the generated image. TYPE: `int`, defaults to `720` DEFAULT: 720
width	The width in pixels of the generated image. TYPE: `int`, defaults to `1280` DEFAULT: 1280
num_frames	The number of frames in the generated video. TYPE: `int`, defaults to `129` DEFAULT: 129
num_inference_steps	The number of denoising steps. More denoising steps usually lead to a higher quality image at the expense of slower inference. TYPE: `int`, defaults to `50` DEFAULT: 50
sigmas	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. TYPE: `List[float]`, *optional* DEFAULT: None
guidance_scale	Guidance scale as defined in Classifier-Free Diffusion Guidance. guidance_scale is defined as w of equation 2. of Imagen Paper. Guidance scale is enabled by setting guidance_scale > 1. Higher guidance scale encourages to generate images that are closely linked to the text prompt, usually at the expense of lower image quality. Note that the only available HunyuanVideo model is CFG-distilled, which means that traditional guidance between unconditional and conditional latent is not applied. TYPE: `float`, defaults to `6.0` DEFAULT: 6.0
num_videos_per_prompt	The number of images to generate per prompt. TYPE: `int`, *optional*, defaults to 1 DEFAULT: 1
generator	A torch.Generator to make generation deterministic. TYPE: `torch.Generator` or `List[torch.Generator]`, *optional* DEFAULT: None
latents	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. TYPE: `torch.Tensor`, *optional* DEFAULT: None
prompt_embeds	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. TYPE: `torch.Tensor`, *optional* DEFAULT: None
output_type	The output format of the generated image. Choose between PIL.Image or np.array. TYPE: `str`, *optional*, defaults to `"pil"` DEFAULT: 'pil'
return_dict	Whether or not to return a [HunyuanVideoPipelineOutput] instead of a plain tuple. TYPE: `bool`, *optional*, defaults to `False` DEFAULT: False
attention_kwargs	A kwargs dictionary that if specified is passed along to the AttentionProcessor as defined under self.processor in diffusers.models.attention_processor. TYPE: `dict`, *optional* DEFAULT: None
clip_skip	Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that the output of the pre-final layer will be used for computing the prompt embeddings. TYPE: `int`, *optional*
callback_on_step_end	A function or a subclass of PipelineCallback or MultiPipelineCallbacks that is called at the end of each denoising step during the inference. 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. TYPE: `Callable`, `PipelineCallback`, `MultiPipelineCallbacks`, *optional* DEFAULT: None
callback_on_step_end_tensor_inputs	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. TYPE: `List`, *optional* DEFAULT: ['latents']

RETURNS	DESCRIPTION
	[~HunyuanVideoPipelineOutput] or tuple: If return_dict is True, [HunyuanVideoPipelineOutput] is returned, otherwise a tuple is returned where the first element is a list with the generated images and the second element is a list of bools indicating whether the corresponding generated image contains "not-safe-for-work" (nsfw) content.

Source code in mindone/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video.py

def __call__(
    self,
    prompt: Union[str, List[str]] = None,
    prompt_2: Union[str, List[str]] = None,
    height: int = 720,
    width: int = 1280,
    num_frames: int = 129,
    num_inference_steps: int = 50,
    sigmas: List[float] = None,
    guidance_scale: float = 6.0,
    num_videos_per_prompt: Optional[int] = 1,
    generator: Optional[Union[np.random.Generator, List[np.random.Generator]]] = None,
    latents: Optional[ms.Tensor] = None,
    prompt_embeds: Optional[ms.Tensor] = None,
    pooled_prompt_embeds: Optional[ms.Tensor] = None,
    prompt_attention_mask: Optional[ms.Tensor] = None,
    output_type: Optional[str] = "pil",
    return_dict: bool = False,
    attention_kwargs: Optional[Dict[str, Any]] = None,
    callback_on_step_end: Optional[
        Union[Callable[[int, int, Dict], None], PipelineCallback, MultiPipelineCallbacks]
    ] = None,
    callback_on_step_end_tensor_inputs: List[str] = ["latents"],
    prompt_template: Dict[str, Any] = DEFAULT_PROMPT_TEMPLATE,
    max_sequence_length: int = 256,
):
    r"""
    The call function to 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.
        height (`int`, defaults to `720`):
            The height in pixels of the generated image.
        width (`int`, defaults to `1280`):
            The width in pixels of the generated image.
        num_frames (`int`, defaults to `129`):
            The number of frames in the generated video.
        num_inference_steps (`int`, 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`, defaults to `6.0`):
            Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
            `guidance_scale` is defined as `w` of equation 2. of [Imagen
            Paper](https://arxiv.org/pdf/2205.11487.pdf). Guidance scale is enabled by setting `guidance_scale >
            1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`,
            usually at the expense of lower image quality. Note that the only available HunyuanVideo model is
            CFG-distilled, which means that traditional guidance between unconditional and conditional latent is
            not applied.
        num_videos_per_prompt (`int`, *optional*, defaults to 1):
            The number of images to generate per prompt.
        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.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.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.
        output_type (`str`, *optional*, defaults to `"pil"`):
            The output format of the generated image. Choose between `PIL.Image` or `np.array`.
        return_dict (`bool`, *optional*, defaults to `False`):
            Whether or not to return a [`HunyuanVideoPipelineOutput`] instead of a plain tuple.
        attention_kwargs (`dict`, *optional*):
            A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
            `self.processor` in
            [diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
        clip_skip (`int`, *optional*):
            Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that
            the output of the pre-final layer will be used for computing the prompt embeddings.
        callback_on_step_end (`Callable`, `PipelineCallback`, `MultiPipelineCallbacks`, *optional*):
            A function or a subclass of `PipelineCallback` or `MultiPipelineCallbacks` that is called at the end of
            each denoising step during the inference. 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.

    Examples:

    Returns:
        [`~HunyuanVideoPipelineOutput`] or `tuple`:
            If `return_dict` is `True`, [`HunyuanVideoPipelineOutput`] is returned, otherwise a `tuple` is returned
            where the first element is a list with the generated images and the second element is a list of `bool`s
            indicating whether the corresponding generated image contains "not-safe-for-work" (nsfw) content.
    """

    if isinstance(callback_on_step_end, (PipelineCallback, MultiPipelineCallbacks)):
        callback_on_step_end_tensor_inputs = callback_on_step_end.tensor_inputs

    # 1. Check inputs. Raise error if not correct
    self.check_inputs(
        prompt,
        prompt_2,
        height,
        width,
        prompt_embeds,
        callback_on_step_end_tensor_inputs,
        prompt_template,
    )

    self._guidance_scale = guidance_scale
    self._attention_kwargs = attention_kwargs
    self._interrupt = False

    # 2. Define call parameters
    if prompt is not None and isinstance(prompt, str):
        batch_size = 1
    elif prompt is not None and isinstance(prompt, list):
        batch_size = len(prompt)
    else:
        batch_size = prompt_embeds.shape[0]

    # 3. Encode input prompt
    prompt_embeds, pooled_prompt_embeds, prompt_attention_mask = self.encode_prompt(
        prompt=prompt,
        prompt_2=prompt_2,
        prompt_template=prompt_template,
        num_videos_per_prompt=num_videos_per_prompt,
        prompt_embeds=prompt_embeds,
        pooled_prompt_embeds=pooled_prompt_embeds,
        prompt_attention_mask=prompt_attention_mask,
        max_sequence_length=max_sequence_length,
    )

    transformer_dtype = self.transformer.dtype
    prompt_embeds = prompt_embeds.to(transformer_dtype)
    prompt_attention_mask = prompt_attention_mask.to(transformer_dtype)
    if pooled_prompt_embeds is not None:
        pooled_prompt_embeds = pooled_prompt_embeds.to(transformer_dtype)

    # 4. Prepare timesteps
    sigmas = np.linspace(1.0, 0.0, num_inference_steps + 1)[:-1] if sigmas is None else sigmas
    timesteps, num_inference_steps = retrieve_timesteps(
        self.scheduler,
        num_inference_steps,
        sigmas=sigmas,
    )

    # 5. Prepare latent variables
    num_channels_latents = self.transformer.config.in_channels
    num_latent_frames = (num_frames - 1) // self.vae_scale_factor_temporal + 1
    latents = self.prepare_latents(
        batch_size * num_videos_per_prompt,
        num_channels_latents,
        height,
        width,
        num_latent_frames,
        ms.float32,
        generator,
        latents,
    )

    # 6. Prepare guidance condition
    guidance = ms.Tensor([guidance_scale] * latents.shape[0], dtype=transformer_dtype) * 1000.0

    # 7. Denoising loop
    num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
    self._num_timesteps = len(timesteps)

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

            latent_model_input = latents.to(transformer_dtype)
            # broadcast to batch dimension in a way that's compatible with ONNX/Core ML
            timestep = t.broadcast_to((latents.shape[0],)).to(latents.dtype)

            noise_pred = self.transformer(
                hidden_states=latent_model_input,
                timestep=timestep,
                encoder_hidden_states=prompt_embeds,
                encoder_attention_mask=prompt_attention_mask,
                pooled_projections=pooled_prompt_embeds,
                guidance=guidance,
                attention_kwargs=attention_kwargs,
                return_dict=False,
            )[0]

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

            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 not output_type == "latent":
        latents = latents.to(self.vae.dtype) / self.vae.config.scaling_factor
        video = self.vae.decode(latents, return_dict=False)[0]
        video = self.video_processor.postprocess_video(video, output_type=output_type)
    else:
        video = latents

    if not return_dict:
        return (video,)

    return HunyuanVideoPipelineOutput(frames=video)

mindone.diffusers.HunyuanVideoPipeline.disable_vae_slicing()

Disable sliced VAE decoding. If enable_vae_slicing was previously enabled, this method will go back to computing decoding in one step.

Source code in mindone/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video.py

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

mindone.diffusers.HunyuanVideoPipeline.disable_vae_tiling()

Disable tiled VAE decoding. If enable_vae_tiling was previously enabled, this method will go back to computing decoding in one step.

Source code in mindone/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video.py

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

mindone.diffusers.HunyuanVideoPipeline.enable_vae_slicing()

Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.

Source code in mindone/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video.py

def enable_vae_slicing(self):
    r"""
    Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
    compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
    """
    self.vae.enable_slicing()

mindone.diffusers.HunyuanVideoPipeline.enable_vae_tiling()

Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow processing larger images.

Source code in mindone/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video.py

def enable_vae_tiling(self):
    r"""
    Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
    compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
    processing larger images.
    """
    self.vae.enable_tiling()

mindone.diffusers.pipelines.hunyuan_video.pipeline_output.HunyuanVideoPipelineOutput dataclass

Bases: BaseOutput

Output class for HunyuanVideo pipelines.

PARAMETER	DESCRIPTION
frames	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). TYPE: `ms.Tensor`, `np.ndarray`, or List[List[PIL.Image.Image]]

Source code in mindone/diffusers/pipelines/hunyuan_video/pipeline_output.py

@dataclass
class HunyuanVideoPipelineOutput(BaseOutput):
    r"""
    Output class for HunyuanVideo pipelines.

    Args:
        frames (`ms.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 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)`.
    """

    frames: ms.Tensor