diffusers/main/id/api/normalization.md

Lapisan normalisasi

Lapisan normalisasi yang dikustomisasi untuk mendukung berbagai model di 🤗 Diffusers.

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

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

diffusers.models.normalization.AdaLayerNorm(embedding_dim: int, num_embeddings: int | None = None, output_dim: int | None = None, norm_elementwise_affine: bool = False, norm_eps: float = 1e-05, chunk_dim: int = 0)

Sumber

Parameter:

embedding_dim (int) : Ukuran setiap vektor embedding.

num_embeddings (int, opsional) : Ukuran dari kamus embedding.

output_dim (int, opsional) --

norm_elementwise_affine (bool, defaultnya `False) --

norm_eps (bool, defaultnya False) --

chunk_dim (int, defaultnya 0) --

Lapisan norm dimodifikasi untuk mengintegrasikan embedding timestep.

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

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

diffusers.models.normalization.AdaLayerNormZero(embedding_dim: int, num_embeddings: int | None = None, norm_type = 'layer_norm', bias = True)

Sumber

Parameter:

embedding_dim (int) : Ukuran setiap vektor embedding.

num_embeddings (int) : Ukuran kamus embeddings.

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

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

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

diffusers.models.normalization.AdaLayerNormSingle(embedding_dim: int, use_additional_conditions: bool = False)

Sumber

Parameter:

embedding_dim (int) : Ukuran setiap vektor embedding.

use_additional_conditions (bool) : Untuk menggunakan kondisi tambahan untuk normalisasi atau tidak.

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

Seperti yang diusulkan dalam PixArt-Alpha (lihat: https://huggingface.co/papers/2310.00426; Bagian 2.3).

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

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

diffusers.models.normalization.AdaGroupNorm(embedding_dim: int, out_dim: int, num_groups: int, act_fn: str | None = None, eps: float = 1e-05)

Sumber

Parameter:

embedding_dim (int) : Ukuran setiap vektor embedding.

num_embeddings (int) : Ukuran kamus embeddings.

num_groups (int) : Jumlah grup untuk memisahkan channel.

act_fn (str, opsional, defaultnya None) : Fungsi aktivasi yang akan digunakan.

eps (float, opsional, defaultnya 1e-5) : Nilai epsilon yang digunakan untuk stabilitas numerik.

Lapisan GroupNorm yang dimodifikasi untuk mengintegrasikan embedding timestep.

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

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

diffusers.models.normalization.AdaLayerNormContinuous(embedding_dim: int, conditioning_embedding_dim: int, elementwise_affine = True, eps = 1e-05, bias = True, norm_type = 'layer_norm')

Sumber

Parameter:

embedding_dim (int) : Dimensi embedding yang digunakan selama proyeksi.

conditioning_embedding_dim (int) : Dimensi dari kondisi input.

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

eps (float, defaultnya 1e-5) : Faktor epsilon.

bias (bias, defaultnya True) : Flag Boolean untuk menunjukkan apakah bias harus digunakan.

norm_type (str, defaultnya "layer_norm") : Lapisan normalisasi yang akan digunakan. Nilai yang didukung: "layer_norm", "rms_norm".

Lapisan normalisasi adaptif dengan lapisan norm (layer_norm atau rms_norm).

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

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

diffusers.models.normalization.RMSNorm(dim, eps: float, elementwise_affine: bool = True, bias: bool = False)

Sumber

Parameter:

dim (int) : Jumlah dimensi yang digunakan untuk weights. Hanya berlaku ketika elementwise_affine bernilai True.

eps (float) : Nilai kecil yang digunakan saat menghitung kebalikan dari akar kuadrat.

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

bias (bool, defaultnya False) : Jika juga melatih param bias.

RMS Norm yang diperkenalkan dalam https://huggingface.co/papers/1910.07467 oleh Zhang dkk.

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

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

diffusers.models.normalization.GlobalResponseNorm(dim)

Sumber

Parameter:

dim (int) : Jumlah dimensi yang digunakan untuk gamma dan beta.

Normalisasi respons global yang diperkenalkan dalam ConvNeXt-v2 (https://huggingface.co/papers/2301.00808).

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

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

diffusers.models.normalization.LuminaLayerNormContinuous(embedding_dim: int, conditioning_embedding_dim: int, elementwise_affine = True, eps = 1e-05, bias = True, norm_type = 'layer_norm', out_dim: int | None = None)

Sumber

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

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

diffusers.models.normalization.SD35AdaLayerNormZeroX(embedding_dim: int, norm_type: str = 'layer_norm', bias: bool = True)

Sumber

Parameter:

embedding_dim (int) : Ukuran setiap vektor embedding.

num_embeddings (int) : Ukuran kamus embeddings.

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

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

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

diffusers.models.normalization.AdaLayerNormZeroSingle(embedding_dim: int, norm_type = 'layer_norm', bias = True)

Sumber

Parameter:

embedding_dim (int) : Ukuran setiap vektor embedding.

num_embeddings (int) : Ukuran kamus embeddings.

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

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

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

diffusers.models.normalization.LuminaRMSNormZero(embedding_dim: int, norm_eps: float, norm_elementwise_affine: bool)

Sumber

Parameter:

embedding_dim (int) : Ukuran setiap vektor embedding.

Norm layer adaptive RMS normalization zero.

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

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

diffusers.models.normalization.LpNorm(p: int = 2, dim: int = -1, eps: float = 1e-12)

Sumber

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

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

diffusers.models.normalization.CogView3PlusAdaLayerNormZeroTextImage(embedding_dim: int, dim: int)

Sumber

Parameter:

embedding_dim (int) : Ukuran setiap vektor embedding.

num_embeddings (int) : Ukuran kamus embeddings.

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

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

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

diffusers.models.normalization.CogVideoXLayerNormZero(conditioning_dim: int, embedding_dim: int, elementwise_affine: bool = True, eps: float = 1e-05, bias: bool = True)

Sumber

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

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

diffusers.models.transformers.transformer_mochi.MochiRMSNormZero(embedding_dim: int, hidden_dim: int, eps: float = 1e-05, elementwise_affine: bool = False)

Sumber

Parameter:

embedding_dim (int) : Ukuran setiap vektor embedding.

Adaptive RMS Norm yang digunakan dalam Mochi.

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

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

diffusers.models.normalization.MochiRMSNorm(dim, eps: float, elementwise_affine: bool = True)

Sumber