if len(contexts)>=2 and the max value the oenultimate list same as of the last list Args: List (_type_): _description_ Returns:

r""" Encodes the prompt into text encoder hidden states. Args: prompt (`str` or `list(int)`): prompt to be encoded

使用含有"{key}" 的模板对多属性文本字典进行拼接，形成新文本 concatenate multiple attribute text dictionaries using a template containing "{key}" to form a new text

使用含有"{}" 的模板对多属性文本元组进行拼接，形成新文本 concatenate multiple attribute text tuples using a template containing "{}" to form a new text Args:

r""" Instantiate a pretrained pytorch model from a pre-trained model configuration. The model is set in evaluation mode by default us

grid_size: int of the grid height and width return: pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or

r""" Returns: `dict` of attention attns: A dictionary containing all attention attns used in the model with indexed by

根据 template_name 确定 prompt 生成器类 choose prompt generator class according to template_name Args: name (str): template 的名字简称，便于指定. te

r""" Prompts can be assigned with local weights using brackets. For example, prompt 'A (very beautiful) masterpiece' highlights the words 'ver

video: b c t1 h w target: b c t2(=1) h w

Loads the ReferenceNet model. Args: sd_referencenet_model (Tuple[str, nn.Module] or str): The pretrained ReferenceNet model or the p

read tensor batch and make a grid with opencv Args: batch (Union[torch.Tensor, np.ndarray]): 4 dim tensor, like b c h w nrows (in

对多属性文本元组进行拼接，形成新文本。 如果`template`含有{key}，则根据key来取值； 如果`template`有且只有1个{}，则根据先后顺序对texts中的值进行拼接。 concatenate multiple attribute

One step for the second-order multistep DPM-Solver. Args: model_output_list (`List[torch.FloatTensor]`):

One step for the third-order multistep DPM-Solver. Args: model_output_list (`List[torch.FloatTensor]`):

Parses a string with attention tokens and returns a list of pairs: text and its associated weight. Accepted tokens are: (abc) - increas

支持多种情况下的latens： img_based_latents: 当Image t=1，latents=None时，使用image赋值到shape，然后加噪；适用于text2video、middle2video。 video_based_late

given a int to represent candidate list, sample n_sample with sample_rate from the candidate list Args: T (int): _description_ n_

save video batch to images into image_type Args: videos (np.array): [h w c] path (str): image directory path

r""" Enable sliced attention computation. When this option is enabled, the attention module will split the input tensor in slices, to

Sets the timesteps used for the diffusion chain. Supporting function to be run before inference. Args: num_inference_ste

_summary_ Args: indexs (List): _description_ n_first (int): _description_ n_last (int): _description_ method (Lit

Predict the sample from the previous timestep by reversing the SDE. This function propagates the diffusion process from the learned m

merge dst into src

使用 lora 更新pipeline中的unet相关参数 Args: pipeline (DiffusionPipeline): _description_ lora (Union[str, Dict]): _description_ alp

use independent ip_adapter attn 中的 to_k, to_v in unet ip_adapter： like ['1.to_k_ip.weight', '1.to_v_ip.weight', '3.to_k_ip.weight'] Args:

r""" 旨在兼容text2video、text2image、img2img、video2video、是否有controlnet等的通用pipeline。目前仅不支持img2img、video2video。 支持多片段同时denoise，交叉部分加权平均

将insight faces 检测的结果转化成prompt convert the results of insight faces detection to prompt Args: face_list (_type_): _desc

有时候一个属性可能会返回多个文本，如 style cartoon会返回宫崎骏和皮克斯两种风格，采用外积增殖成多个字典。 sometimes an attribute may return multiple texts, such as style cartoon w