From f009f3bcc47e0e43f70a1ead20dae936ef8d01f1 Mon Sep 17 00:00:00 2001 From: taoyuan-guo Date: Thu, 9 Jan 2025 21:16:56 +0800 Subject: [PATCH] huanyuanvideo pipeline --- .../hyvideo/pipelines/__init__.py | 18 + .../pipelines/pipeline_hunyuan_video.py | 903 ++++++++++++++++++ 2 files changed, 921 insertions(+) create mode 100644 MindIE/MindIE-Torch/built-in/foundation/hunyuan_video/hyvideo/pipelines/__init__.py create mode 100644 MindIE/MindIE-Torch/built-in/foundation/hunyuan_video/hyvideo/pipelines/pipeline_hunyuan_video.py diff --git a/MindIE/MindIE-Torch/built-in/foundation/hunyuan_video/hyvideo/pipelines/__init__.py b/MindIE/MindIE-Torch/built-in/foundation/hunyuan_video/hyvideo/pipelines/__init__.py new file mode 100644 index 0000000000..e3930addd7 --- /dev/null +++ b/MindIE/MindIE-Torch/built-in/foundation/hunyuan_video/hyvideo/pipelines/__init__.py @@ -0,0 +1,18 @@ +#!/usr/bin/env python +# coding=utf-8 +# Copyright 2024 Huawei Technologies Co., Ltd +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + + +from .pipeline_hunyuan_video import HunyuanVideoPipeline \ No newline at end of file diff --git a/MindIE/MindIE-Torch/built-in/foundation/hunyuan_video/hyvideo/pipelines/pipeline_hunyuan_video.py b/MindIE/MindIE-Torch/built-in/foundation/hunyuan_video/hyvideo/pipelines/pipeline_hunyuan_video.py new file mode 100644 index 0000000000..52f716faa1 --- /dev/null +++ b/MindIE/MindIE-Torch/built-in/foundation/hunyuan_video/hyvideo/pipelines/pipeline_hunyuan_video.py @@ -0,0 +1,903 @@ +#!/usr/bin/env python +# coding=utf-8 +# Copyright 2024 Huawei Technologies Co., Ltd +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + + +import inspect +from typing import Any, Callable, Dict, List, Optional, Union, Tuple +import torch +import torch_npu +import torch.distributed as dist +import numpy as np +from dataclasses import dataclass +from packaging import version + +from diffusers.callbacks import MultiPipelineCallbacks, PipelineCallback +from diffusers.configuration_utils import FrozenDict +from diffusers.image_processor import VaeImageProcessor +from diffusers.loaders import LoraLoaderMixin, TextualInversionLoaderMixin +from diffusers.models import AutoencoderKL +from diffusers.models.lora import adjust_lora_scale_text_encoder +from diffusers.schedulers import KarrasDiffusionSchedulers +from diffusers.utils import ( + USE_PEFT_BACKEND, + deprecate, + logging, + replace_example_docstring, + scale_lora_layers, + unscale_lora_layers, +) +from diffusers.utils.torch_utils import randn_tensor +from diffusers.pipelines.pipeline_utils import DiffusionPipeline +from diffusers.utils import BaseOutput + +from ...constants import PRECISION_TO_TYPE +from ...vae.autoencoder_kl_causal_3d import AutoencoderKLCausal3D +from ...text_encoder import TextEncoder +from ...modules import HYVideoDiffusionTransformer + +logger = logging.get_logger(__name__) # pylint: disable=invalid-name + +EXAMPLE_DOC_STRING = """""" + + +def rescale_noise_cfg(noise_cfg, noise_pred_text, guidance_rescale=0.0): + """ + Rescale `noise_cfg` according to `guidance_rescale`. Based on findings of [Common Diffusion Noise Schedules and + Sample Steps are Flawed](https://arxiv.org/pdf/2305.08891.pdf). See Section 3.4 + """ + std_text = noise_pred_text.std( + dim=list(range(1, noise_pred_text.ndim)), keepdim=True + ) + std_cfg = noise_cfg.std(dim=list(range(1, noise_cfg.ndim)), keepdim=True) + # rescale the results from guidance (fixes overexposure) + noise_pred_rescaled = noise_cfg * (std_text / std_cfg) + # mix with the original results from guidance by factor guidance_rescale to avoid "plain looking" images + noise_cfg = ( + guidance_rescale * noise_pred_rescaled + (1 - guidance_rescale) * noise_cfg + ) + return noise_cfg + + +def retrieve_timesteps( + scheduler, + num_inference_steps: Optional[int] = None, + device: Optional[Union[str, torch.device]] = None, + timesteps: Optional[List[int]] = None, + sigmas: Optional[List[float]] = None, + **kwargs, +): + """ + Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles + custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`. + + Args: + 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`. + 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 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: + raise ValueError( + f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" + f" timestep schedules. Please check whether you are using the correct scheduler." + ) + scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs) + timesteps = scheduler.timesteps + num_inference_steps = len(timesteps) + elif sigmas is not None: + accept_sigmas = "sigmas" in set( + inspect.signature(scheduler.set_timesteps).parameters.keys() + ) + if not accept_sigmas: + raise ValueError( + f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom" + f" sigmas schedules. Please check whether you are using the correct scheduler." + ) + scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs) + timesteps = scheduler.timesteps + num_inference_steps = len(timesteps) + else: + scheduler.set_timesteps(num_inference_steps, device=device, **kwargs) + timesteps = scheduler.timesteps + return timesteps, num_inference_steps + + +@dataclass +class HunyuanVideoPipelineOutput(BaseOutput): + videos: Union[torch.Tensor, np.ndarray] + + +class HunyuanVideoPipeline(DiffusionPipeline): + 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: + vae ([`AutoencoderKL`]): + Variational Auto-Encoder (VAE) model to encode and decode images to and from latent representations. + text_encoder ([`TextEncoder`]): + Frozen text-encoder. + text_encoder_2 ([`TextEncoder`]): + Frozen text-encoder_2. + transformer ([`HYVideoDiffusionTransformer`]): + A `HYVideoDiffusionTransformer` to denoise the encoded video latents. + scheduler ([`SchedulerMixin`]): + A scheduler to be used in combination with `unet` to denoise the encoded image latents. + """ + + model_cpu_offload_seq = "text_encoder->text_encoder_2->transformer->vae" + _optional_components = ["text_encoder_2"] + _exclude_from_cpu_offload = ["transformer"] + _callback_tensor_inputs = ["latents", "prompt_embeds", "negative_prompt_embeds"] + + def __init__( + self, + vae: AutoencoderKL, + text_encoder: TextEncoder, + transformer: HYVideoDiffusionTransformer, + scheduler: KarrasDiffusionSchedulers, + text_encoder_2: Optional[TextEncoder] = None, + progress_bar_config: Dict[str, Any] = None, + args=None, + ): + super().__init__() + + # ========================================================================================== + if progress_bar_config is None: + progress_bar_config = {} + if not hasattr(self, "_progress_bar_config"): + self._progress_bar_config = {} + self._progress_bar_config.update(progress_bar_config) + + self.args = args + # ========================================================================================== + + if ( + hasattr(scheduler.config, "steps_offset") + and scheduler.config.steps_offset != 1 + ): + deprecation_message = ( + f"The configuration file of this scheduler: {scheduler} is outdated. `steps_offset`" + f" should be set to 1 instead of {scheduler.config.steps_offset}. Please make sure " + "to update the config accordingly as leaving `steps_offset` might led to incorrect results" + " in future versions. If you have downloaded this checkpoint from the Hugging Face Hub," + " it would be very nice if you could open a Pull request for the `scheduler/scheduler_config.json`" + " file" + ) + deprecate( + "steps_offset!=1", "1.0.0", deprecation_message, standard_warn=False + ) + new_config = dict(scheduler.config) + new_config["steps_offset"] = 1 + scheduler._internal_dict = FrozenDict(new_config) + + if ( + hasattr(scheduler.config, "clip_sample") + and scheduler.config.clip_sample is True + ): + deprecation_message = ( + f"The configuration file of this scheduler: {scheduler} has not set the configuration `clip_sample`." + " `clip_sample` should be set to False in the configuration file. Please make sure to update the" + " config accordingly as not setting `clip_sample` in the config might lead to incorrect results in" + " future versions. If you have downloaded this checkpoint from the Hugging Face Hub, it would be very" + " nice if you could open a Pull request for the `scheduler/scheduler_config.json` file" + ) + deprecate( + "clip_sample not set", "1.0.0", deprecation_message, standard_warn=False + ) + new_config = dict(scheduler.config) + new_config["clip_sample"] = False + scheduler._internal_dict = FrozenDict(new_config) + + self.register_modules( + vae=vae, + text_encoder=text_encoder, + transformer=transformer, + scheduler=scheduler, + text_encoder_2=text_encoder_2, + ) + self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) + self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor) + + def encode_prompt( + self, + prompt, + device, + num_videos_per_prompt, + do_classifier_free_guidance, + negative_prompt=None, + prompt_embeds: Optional[torch.Tensor] = None, + attention_mask: Optional[torch.Tensor] = None, + negative_prompt_embeds: Optional[torch.Tensor] = None, + negative_attention_mask: Optional[torch.Tensor] = None, + lora_scale: Optional[float] = None, + clip_skip: Optional[int] = None, + text_encoder: Optional[TextEncoder] = None, + data_type: Optional[str] = "image", + ): + r""" + Encodes the prompt into text encoder hidden states. + + Args: + prompt (`str` or `List[str]`, *optional*): + prompt to be encoded + device: (`torch.device`): + torch device + num_videos_per_prompt (`int`): + number of videos that should be generated per prompt + do_classifier_free_guidance (`bool`): + whether to use classifier free guidance or not + negative_prompt (`str` or `List[str]`, *optional*): + The prompt or prompts not to guide the video 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.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. + attention_mask (`torch.Tensor`, *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. + negative_attention_mask (`torch.Tensor`, *optional*): + lora_scale (`float`, *optional*): + A LoRA scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded. + 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. + text_encoder (TextEncoder, *optional*): + data_type (`str`, *optional*): + """ + if text_encoder is None: + text_encoder = self.text_encoder + + # set lora scale so that monkey patched LoRA + # function of text encoder can correctly access it + if lora_scale is not None and isinstance(self, LoraLoaderMixin): + self._lora_scale = lora_scale + + # dynamically adjust the LoRA scale + if not USE_PEFT_BACKEND: + adjust_lora_scale_text_encoder(text_encoder.model, lora_scale) + else: + scale_lora_layers(text_encoder.model, lora_scale) + + if prompt is not None and isinstance(prompt, str): + batch_size = 1 + elif prompt is not None and isinstance(prompt, list): + batch_size = len(prompt) + else: + batch_size = prompt_embeds.shape[0] + + if prompt_embeds is None: + # textual inversion: process multi-vector tokens if necessary + if isinstance(self, TextualInversionLoaderMixin): + prompt = self.maybe_convert_prompt(prompt, text_encoder.tokenizer) + + text_inputs = text_encoder.text2tokens(prompt, data_type=data_type) + + if clip_skip is None: + prompt_outputs = text_encoder.encode( + text_inputs, data_type=data_type, device=device + ) + prompt_embeds = prompt_outputs.hidden_state + else: + prompt_outputs = text_encoder.encode( + text_inputs, + output_hidden_states=True, + data_type=data_type, + device=device, + ) + # Access the `hidden_states` first, that contains a tuple of + # all the hidden states from the encoder layers. Then index into + # the tuple to access the hidden states from the desired layer. + prompt_embeds = prompt_outputs.hidden_states_list[-(clip_skip + 1)] + # We also need to apply the final LayerNorm here to not mess with the + # representations. The `last_hidden_states` that we typically use for + # obtaining the final prompt representations passes through the LayerNorm + # layer. + prompt_embeds = text_encoder.model.text_model.final_layer_norm( + prompt_embeds + ) + + attention_mask = prompt_outputs.attention_mask + if attention_mask is not None: + attention_mask = attention_mask.to(device) + bs_embed, seq_len = attention_mask.shape + attention_mask = attention_mask.repeat(1, num_videos_per_prompt) + attention_mask = attention_mask.view( + bs_embed * num_videos_per_prompt, seq_len + ) + + if text_encoder is not None: + prompt_embeds_dtype = text_encoder.dtype + elif self.transformer is not None: + prompt_embeds_dtype = self.transformer.dtype + else: + prompt_embeds_dtype = prompt_embeds.dtype + + prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device) + + if prompt_embeds.ndim == 2: + bs_embed, _ = prompt_embeds.shape + # duplicate text embeddings for each generation per prompt, using mps friendly method + prompt_embeds = prompt_embeds.repeat(1, num_videos_per_prompt) + prompt_embeds = prompt_embeds.view(bs_embed * num_videos_per_prompt, -1) + else: + bs_embed, seq_len, _ = prompt_embeds.shape + # duplicate text embeddings for each generation per prompt, using mps friendly method + prompt_embeds = prompt_embeds.repeat(1, num_videos_per_prompt, 1) + prompt_embeds = prompt_embeds.view( + bs_embed * num_videos_per_prompt, seq_len, -1 + ) + + # get unconditional embeddings for classifier free guidance + if do_classifier_free_guidance and negative_prompt_embeds is None: + uncond_tokens: List[str] + if negative_prompt is None: + uncond_tokens = [""] * batch_size + elif prompt is not None and type(prompt) is not type(negative_prompt): + raise TypeError( + f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !=" + f" {type(prompt)}." + ) + elif isinstance(negative_prompt, str): + uncond_tokens = [negative_prompt] + elif batch_size != len(negative_prompt): + raise ValueError( + f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:" + f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches" + " the batch size of `prompt`." + ) + else: + uncond_tokens = negative_prompt + + # textual inversion: process multi-vector tokens if necessary + if isinstance(self, TextualInversionLoaderMixin): + uncond_tokens = self.maybe_convert_prompt( + uncond_tokens, text_encoder.tokenizer + ) + + # max_length = prompt_embeds.shape[1] + uncond_input = text_encoder.text2tokens(uncond_tokens, data_type=data_type) + + negative_prompt_outputs = text_encoder.encode( + uncond_input, data_type=data_type, device=device + ) + negative_prompt_embeds = negative_prompt_outputs.hidden_state + + negative_attention_mask = negative_prompt_outputs.attention_mask + if negative_attention_mask is not None: + negative_attention_mask = negative_attention_mask.to(device) + _, seq_len = negative_attention_mask.shape + negative_attention_mask = negative_attention_mask.repeat( + 1, num_videos_per_prompt + ) + negative_attention_mask = negative_attention_mask.view( + batch_size * num_videos_per_prompt, seq_len + ) + + 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=prompt_embeds_dtype, device=device + ) + + if negative_prompt_embeds.ndim == 2: + negative_prompt_embeds = negative_prompt_embeds.repeat( + 1, num_videos_per_prompt + ) + negative_prompt_embeds = negative_prompt_embeds.view( + batch_size * num_videos_per_prompt, -1 + ) + else: + negative_prompt_embeds = negative_prompt_embeds.repeat( + 1, num_videos_per_prompt, 1 + ) + negative_prompt_embeds = negative_prompt_embeds.view( + batch_size * num_videos_per_prompt, seq_len, -1 + ) + + if text_encoder is not None: + if isinstance(self, LoraLoaderMixin) and USE_PEFT_BACKEND: + # Retrieve the original scale by scaling back the LoRA layers + unscale_lora_layers(text_encoder.model, lora_scale) + + return ( + prompt_embeds, + negative_prompt_embeds, + attention_mask, + negative_attention_mask, + ) + + @torch.no_grad() + @replace_example_docstring(EXAMPLE_DOC_STRING) + def __call__( + self, + prompt: Union[str, List[str]], + height: int, + width: int, + video_length: int, + data_type: str = "video", + num_inference_steps: int = 50, + timesteps: List[int] = None, + sigmas: List[float] = None, + guidance_scale: float = 7.5, + 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.Tensor] = None, + prompt_embeds: Optional[torch.Tensor] = None, + attention_mask: Optional[torch.Tensor] = None, + negative_prompt_embeds: Optional[torch.Tensor] = None, + negative_attention_mask: Optional[torch.Tensor] = None, + output_type: Optional[str] = "pil", + return_dict: bool = True, + cross_attention_kwargs: Optional[Dict[str, Any]] = None, + guidance_rescale: float = 0.0, + clip_skip: Optional[int] = None, + callback_on_step_end: Optional[ + Union[ + Callable[[int, int, Dict], None], + PipelineCallback, + MultiPipelineCallbacks, + ] + ] = None, + callback_on_step_end_tensor_inputs: List[str] = ["latents"], + freqs_cis: Tuple[torch.Tensor, torch.Tensor] = None, + vae_ver: str = "88-4c-sd", + enable_tiling: bool = False, + n_tokens: Optional[int] = None, + embedded_guidance_scale: Optional[float] = None, + **kwargs, + ): + r""" + The call function to the pipeline for generation. + + Args: + prompt (`str` or `List[str]`): + The prompt or prompts to guide image generation. If not defined, you need to pass `prompt_embeds`. + height (`int`): + The height in pixels of the generated image. + width (`int`): + The width in pixels of the generated image. + video_length (`int`): + The number of frames in the generated video. + 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. + 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. + guidance_scale (`float`, *optional*, defaults to 7.5): + A higher guidance scale value encourages the model to generate images closely linked to the text + `prompt` at the expense of lower image quality. Guidance scale is enabled when `guidance_scale > 1`. + negative_prompt (`str` or `List[str]`, *optional*): + The prompt or prompts to guide what to not include in image generation. If not defined, you need to + pass `negative_prompt_embeds` instead. Ignored when not using guidance (`guidance_scale < 1`). + num_videos_per_prompt (`int`, *optional*, defaults to 1): + The number of images to generate per prompt. + eta (`float`, *optional*, defaults to 0.0): + Corresponds to parameter eta (η) from the [DDIM](https://arxiv.org/abs/2010.02502) paper. Only applies + to the [`~schedulers.DDIMScheduler`], and is ignored in other schedulers. + 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. + 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. + + 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 `True`): + Whether or not to return a [`HunyuanVideoPipelineOutput`] 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). + guidance_rescale (`float`, *optional*, defaults to 0.0): + Guidance rescale factor from [Common Diffusion Noise Schedules and Sample Steps are + Flawed](https://arxiv.org/pdf/2305.08891.pdf). Guidance rescale factor should fix overexposure when + using zero terminal SNR. + 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. + """ + callback = kwargs.pop("callback", None) + callback_steps = kwargs.pop("callback_steps", None) + + if callback is not None: + deprecate( + "callback", + "1.0.0", + "Passing `callback` as an input argument to `__call__` is deprecated, consider using `callback_on_step_end`", + ) + if callback_steps is not None: + deprecate( + "callback_steps", + "1.0.0", + "Passing `callback_steps` as an input argument to `__call__` is deprecated, consider using `callback_on_step_end`", + ) + + if isinstance(callback_on_step_end, (PipelineCallback, MultiPipelineCallbacks)): + callback_on_step_end_tensor_inputs = callback_on_step_end.tensor_inputs + + # 0. Default height and width to unet + # height = height or self.transformer.config.sample_size * self.vae_scale_factor + # width = width or self.transformer.config.sample_size * self.vae_scale_factor + # to deal with lora scaling and other possible forward hooks + + # 1. Check inputs. Raise error if not correct + self.check_inputs( + prompt, + height, + width, + video_length, + callback_steps, + negative_prompt, + prompt_embeds, + negative_prompt_embeds, + callback_on_step_end_tensor_inputs, + vae_ver=vae_ver, + ) + + self._guidance_scale = guidance_scale + self._guidance_rescale = guidance_rescale + self._clip_skip = clip_skip + self._cross_attention_kwargs = cross_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] + + device = torch.device(f"cuda:{dist.get_rank()}") if dist.is_initialized() else self._execution_device + stream = torch_npu.npu.Stream() + + # 3. Encode input prompt + lora_scale = ( + self.cross_attention_kwargs.get("scale", None) + if self.cross_attention_kwargs is not None + else None + ) + + with torch_npu.npu.stream(stream): + if next(self.text_encoder.parameters()).device == torch.device("cpu"): + self.text_encoder.to(device) + if next(self.text_encoder_2.parameters()).device == torch.device("cpu"): + self.text_encoder_2.to(device) + + ( + prompt_embeds, + negative_prompt_embeds, + prompt_mask, + negative_prompt_mask, + ) = self.encode_prompt( + prompt, + device, + num_videos_per_prompt, + self.do_classifier_free_guidance, + negative_prompt, + prompt_embeds=prompt_embeds, + attention_mask=attention_mask, + negative_prompt_embeds=negative_prompt_embeds, + negative_attention_mask=negative_attention_mask, + lora_scale=lora_scale, + clip_skip=self.clip_skip, + data_type=data_type, + ) + if self.text_encoder_2 is not None: + ( + prompt_embeds_2, + negative_prompt_embeds_2, + prompt_mask_2, + negative_prompt_mask_2, + ) = self.encode_prompt( + prompt, + device, + num_videos_per_prompt, + self.do_classifier_free_guidance, + negative_prompt, + prompt_embeds=None, + attention_mask=None, + negative_prompt_embeds=None, + negative_attention_mask=None, + lora_scale=lora_scale, + clip_skip=self.clip_skip, + text_encoder=self.text_encoder_2, + data_type=data_type, + ) + else: + prompt_embeds_2 = None + negative_prompt_embeds_2 = None + prompt_mask_2 = None + negative_prompt_mask_2 = None + + with torch_npu.npu.stream(stream): + if self.text_encoder.parameters().device != torch.device("cpu"): + self.text_encoder.to("cpu") + if self.text_encoder_2.parameters().device != torch.device("cpu"): + self.text_encoder_2.to("cpu") + torch.npu.empty_cache() + + # For classifier free guidance, we need to do two forward passes. + # Here we concatenate the unconditional and text embeddings into a single batch + # to avoid doing two forward passes + if self.do_classifier_free_guidance: + prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds]) + if prompt_mask is not None: + prompt_mask = torch.cat([negative_prompt_mask, prompt_mask]) + if prompt_embeds_2 is not None: + prompt_embeds_2 = torch.cat([negative_prompt_embeds_2, prompt_embeds_2]) + if prompt_mask_2 is not None: + prompt_mask_2 = torch.cat([negative_prompt_mask_2, prompt_mask_2]) + + + # 4. Prepare timesteps + extra_set_timesteps_kwargs = self.prepare_extra_func_kwargs( + self.scheduler.set_timesteps, {"n_tokens": n_tokens} + ) + timesteps, num_inference_steps = retrieve_timesteps( + self.scheduler, + num_inference_steps, + device, + timesteps, + sigmas, + **extra_set_timesteps_kwargs, + ) + + if "884" in vae_ver: + video_length = (video_length - 1) // 4 + 1 + elif "888" in vae_ver: + video_length = (video_length - 1) // 8 + 1 + else: + video_length = video_length + + # 5. Prepare latent variables + num_channels_latents = self.transformer.config.in_channels + latents = self.prepare_latents( + batch_size * num_videos_per_prompt, + num_channels_latents, + height, + width, + video_length, + prompt_embeds.dtype, + device, + generator, + latents, + ) + + # 6. Prepare extra step kwargs. TODO: Logic should ideally just be moved out of the pipeline + extra_step_kwargs = self.prepare_extra_func_kwargs( + self.scheduler.step, + {"generator": generator, "eta": eta}, + ) + + target_dtype = PRECISION_TO_TYPE[self.args.precision] + autocast_enabled = ( + target_dtype != torch.float32 + ) and not self.args.disable_autocast + vae_dtype = PRECISION_TO_TYPE[self.args.vae_precision] + vae_autocast_enabled = ( + vae_dtype != torch.float32 + ) and not self.args.disable_autocast + + # 7. Denoising loop + num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order + self._num_timesteps = len(timesteps) + + # if is_progress_bar: + with self.progress_bar(total=num_inference_steps) as progress_bar: + for i, t in enumerate(timesteps): + if self.interrupt: + continue + + # 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 + ) + latent_model_input = self.scheduler.scale_model_input( + latent_model_input, t + ) + + t_expand = t.repeat(latent_model_input.shape[0]) + guidance_expand = ( + torch.tensor( + [embedded_guidance_scale] * latent_model_input.shape[0], + dtype=torch.float32, + device=device, + ).to(target_dtype) + * 1000.0 + if embedded_guidance_scale is not None + else None + ) + + # predict the noise residual + with torch.autocast( + device_type="cuda", dtype=target_dtype, enabled=autocast_enabled + ): + noise_pred = self.transformer( # For an input image (129, 192, 336) (1, 256, 256) + latent_model_input, # [2, 16, 33, 24, 42] + t_expand, # [2] + text_states=prompt_embeds, # [2, 256, 4096] + text_mask=prompt_mask, # [2, 256] + text_states_2=prompt_embeds_2, # [2, 768] + freqs_cos=freqs_cis[0], # [seqlen, head_dim] + freqs_sin=freqs_cis[1], # [seqlen, head_dim] + guidance=guidance_expand, + return_dict=True, + )[ + "x" + ] + + # perform guidance + if self.do_classifier_free_guidance: + noise_pred_uncond, noise_pred_text = noise_pred.chunk(2) + noise_pred = noise_pred_uncond + self.guidance_scale * ( + noise_pred_text - noise_pred_uncond + ) + + if self.do_classifier_free_guidance and self.guidance_rescale > 0.0: + # Based on 3.4. in https://arxiv.org/pdf/2305.08891.pdf + noise_pred = rescale_noise_cfg( + noise_pred, + noise_pred_text, + guidance_rescale=self.guidance_rescale, + ) + + # compute the previous noisy sample x_t -> x_t-1 + latents = self.scheduler.step( + noise_pred, t, latents, **extra_step_kwargs, 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) + negative_prompt_embeds = callback_outputs.pop( + "negative_prompt_embeds", negative_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 + ): + if progress_bar is not None: + progress_bar.update() + if callback is not None and i % callback_steps == 0: + step_idx = i // getattr(self.scheduler, "order", 1) + callback(step_idx, t, latents) + + if not output_type == "latent": + expand_temporal_dim = False + if len(latents.shape) == 4: + if isinstance(self.vae, AutoencoderKLCausal3D): + latents = latents.unsqueeze(2) + expand_temporal_dim = True + elif len(latents.shape) == 5: + pass + else: + raise ValueError( + f"Only support latents with shape (b, c, h, w) or (b, c, f, h, w), but got {latents.shape}." + ) + + if ( + hasattr(self.vae.config, "shift_factor") + and self.vae.config.shift_factor + ): + latents = ( + latents / self.vae.config.scaling_factor + + self.vae.config.shift_factor + ) + else: + latents = latents / self.vae.config.scaling_factor + + with torch.autocast( + device_type="cuda", dtype=vae_dtype, enabled=vae_autocast_enabled + ): + if enable_tiling: + self.vae.enable_tiling() + image = self.vae.decode( + latents, return_dict=False, generator=generator + )[0] + else: + image = self.vae.decode( + latents, return_dict=False, generator=generator + )[0] + + if expand_temporal_dim or image.shape[2] == 1: + image = image.squeeze(2) + + else: + image = latents + + image = (image / 2 + 0.5).clamp(0, 1) + # we always cast to float32 as this does not cause significant overhead and is compatible with bfloa16 + image = image.cpu().float() + + # Offload all models + self.maybe_free_model_hooks() + + if not return_dict: + return image + + return HunyuanVideoPipelineOutput(videos=image) \ No newline at end of file -- Gitee