diffusers/pr_12448/en/api/normalization.md

Normalization layers

Customized normalization layers for supporting various models in 🤗 Diffusers.

AdaLayerNorm[[diffusers.models.normalization.AdaLayerNorm]]

diffusers.models.normalization.AdaLayerNorm[[diffusers.models.normalization.AdaLayerNorm]]

Source

Norm layer modified to incorporate timestep embeddings.

Parameters:

embedding_dim (int) : The size of each embedding vector.

num_embeddings (int, optional) : The size of the embeddings dictionary.

output_dim (int, optional) --

norm_elementwise_affine (bool, defaults to `False) --

norm_eps (bool, defaults to False) --

chunk_dim (int, defaults to 0) --

AdaLayerNormZero[[diffusers.models.normalization.AdaLayerNormZero]]

diffusers.models.normalization.AdaLayerNormZero[[diffusers.models.normalization.AdaLayerNormZero]]

Source

Norm layer adaptive layer norm zero (adaLN-Zero).

Parameters:

embedding_dim (int) : The size of each embedding vector.

num_embeddings (int) : The size of the embeddings dictionary.

AdaLayerNormSingle[[diffusers.models.normalization.AdaLayerNormSingle]]

diffusers.models.normalization.AdaLayerNormSingle[[diffusers.models.normalization.AdaLayerNormSingle]]

Source

Norm layer adaptive layer norm single (adaLN-single).

As proposed in PixArt-Alpha (see: https://huggingface.co/papers/2310.00426; Section 2.3).

Parameters:

embedding_dim (int) : The size of each embedding vector.

use_additional_conditions (bool) : To use additional conditions for normalization or not.

AdaGroupNorm[[diffusers.models.normalization.AdaGroupNorm]]

diffusers.models.normalization.AdaGroupNorm[[diffusers.models.normalization.AdaGroupNorm]]

Source

GroupNorm layer modified to incorporate timestep embeddings.

Parameters:

embedding_dim (int) : The size of each embedding vector.

num_embeddings (int) : The size of the embeddings dictionary.

num_groups (int) : The number of groups to separate the channels into.

act_fn (str, optional, defaults to None) : The activation function to use.

eps (float, optional, defaults to 1e-5) : The epsilon value to use for numerical stability.

AdaLayerNormContinuous[[diffusers.models.normalization.AdaLayerNormContinuous]]

diffusers.models.normalization.AdaLayerNormContinuous[[diffusers.models.normalization.AdaLayerNormContinuous]]

Source

Adaptive normalization layer with a norm layer (layer_norm or rms_norm).

Parameters:

embedding_dim (int) : Embedding dimension to use during projection.

conditioning_embedding_dim (int) : Dimension of the input condition.

elementwise_affine (bool, defaults to True) : Boolean flag to denote if affine transformation should be applied.

eps (float, defaults to 1e-5) : Epsilon factor.

bias (bias, defaults to True) : Boolean flag to denote if bias should be use.

norm_type (str, defaults to "layer_norm") : Normalization layer to use. Values supported: "layer_norm", "rms_norm".

RMSNorm[[diffusers.models.normalization.RMSNorm]]

diffusers.models.normalization.RMSNorm[[diffusers.models.normalization.RMSNorm]]

Source

RMS Norm as introduced in https://huggingface.co/papers/1910.07467 by Zhang et al.

Parameters:

dim (int) : Number of dimensions to use for weights. Only effective when elementwise_affine is True.

eps (float) : Small value to use when calculating the reciprocal of the square-root.

elementwise_affine (bool, defaults to True) : Boolean flag to denote if affine transformation should be applied.

bias (bool, defaults to False) : If also training the bias param.

GlobalResponseNorm[[diffusers.models.normalization.GlobalResponseNorm]]

diffusers.models.normalization.GlobalResponseNorm[[diffusers.models.normalization.GlobalResponseNorm]]

Source

Global response normalization as introduced in ConvNeXt-v2 (https://huggingface.co/papers/2301.00808).

Parameters:

dim (int) : Number of dimensions to use for the gamma and beta.

LuminaLayerNormContinuous[[diffusers.models.normalization.LuminaLayerNormContinuous]]

diffusers.models.normalization.LuminaLayerNormContinuous[[diffusers.models.normalization.LuminaLayerNormContinuous]]

Source

SD35AdaLayerNormZeroX[[diffusers.models.normalization.SD35AdaLayerNormZeroX]]

diffusers.models.normalization.SD35AdaLayerNormZeroX[[diffusers.models.normalization.SD35AdaLayerNormZeroX]]

Source

Norm layer adaptive layer norm zero (AdaLN-Zero).

Parameters:

embedding_dim (int) : The size of each embedding vector.

num_embeddings (int) : The size of the embeddings dictionary.

AdaLayerNormZeroSingle[[diffusers.models.normalization.AdaLayerNormZeroSingle]]

diffusers.models.normalization.AdaLayerNormZeroSingle[[diffusers.models.normalization.AdaLayerNormZeroSingle]]

Source

Norm layer adaptive layer norm zero (adaLN-Zero).

Parameters:

embedding_dim (int) : The size of each embedding vector.

num_embeddings (int) : The size of the embeddings dictionary.

LuminaRMSNormZero[[diffusers.models.normalization.LuminaRMSNormZero]]

diffusers.models.normalization.LuminaRMSNormZero[[diffusers.models.normalization.LuminaRMSNormZero]]

Source

Norm layer adaptive RMS normalization zero.

Parameters:

embedding_dim (int) : The size of each embedding vector.

LpNorm[[diffusers.models.normalization.LpNorm]]

diffusers.models.normalization.LpNorm[[diffusers.models.normalization.LpNorm]]

Source

CogView3PlusAdaLayerNormZeroTextImage[[diffusers.models.normalization.CogView3PlusAdaLayerNormZeroTextImage]]

diffusers.models.normalization.CogView3PlusAdaLayerNormZeroTextImage[[diffusers.models.normalization.CogView3PlusAdaLayerNormZeroTextImage]]

Source

Norm layer adaptive layer norm zero (adaLN-Zero).

Parameters:

embedding_dim (int) : The size of each embedding vector.

num_embeddings (int) : The size of the embeddings dictionary.

CogVideoXLayerNormZero[[diffusers.models.normalization.CogVideoXLayerNormZero]]

diffusers.models.normalization.CogVideoXLayerNormZero[[diffusers.models.normalization.CogVideoXLayerNormZero]]

Source

MochiRMSNormZero[[diffusers.models.transformers.transformer_mochi.MochiRMSNormZero]]

diffusers.models.transformers.transformer_mochi.MochiRMSNormZero[[diffusers.models.transformers.transformer_mochi.MochiRMSNormZero]]

Source

Adaptive RMS Norm used in Mochi.

Parameters:

embedding_dim (int) : The size of each embedding vector.

MochiRMSNorm[[diffusers.models.normalization.MochiRMSNorm]]

diffusers.models.normalization.MochiRMSNorm[[diffusers.models.normalization.MochiRMSNorm]]

Source