Hugging Face's logo

Diffusers documentation

Attention Processor

Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

Attention Processor

An attention processor is a class for applying different types of attention mechanisms.

AttnProcessor

class diffusers.models.attention_processor.AttnProcessor

< >

( )

Default processor for performing attention-related computations.

class diffusers.models.attention_processor.AttnProcessor2_0

< >

( )

Processor for implementing scaled dot-product attention (enabled by default if you’re using PyTorch 2.0).

class diffusers.models.attention_processor.AttnAddedKVProcessor

< >

( )

Processor for performing attention-related computations with extra learnable key and value matrices for the text encoder.

class diffusers.models.attention_processor.AttnAddedKVProcessor2_0

< >

( )

Processor for performing scaled dot-product attention (enabled by default if you’re using PyTorch 2.0), with extra learnable key and value matrices for the text encoder.

class diffusers.models.attention_processor.AttnProcessorNPU

< >

( )

Processor for implementing flash attention using torch_npu. Torch_npu supports only fp16 and bf16 data types. If fp32 is used, F.scaled_dot_product_attention will be used for computation, but the acceleration effect on NPU is not significant.

class diffusers.models.attention_processor.FusedAttnProcessor2_0

< >

( )

Processor for implementing scaled dot-product attention (enabled by default if you’re using PyTorch 2.0). It uses fused projection layers. For self-attention modules, all projection matrices (i.e., query, key, value) are fused. For cross-attention modules, key and value projection matrices are fused.

This API is currently 🧪 experimental in nature and can change in future.

Allegro

class diffusers.models.attention_processor.AllegroAttnProcessor2_0

< >

( )

Processor for implementing scaled dot-product attention (enabled by default if you’re using PyTorch 2.0). This is used in the Allegro model. It applies a normalization layer and rotary embedding on the query and key vector.

AuraFlow

class diffusers.models.attention_processor.AuraFlowAttnProcessor2_0

< >

( )

Attention processor used typically in processing Aura Flow.

class diffusers.models.attention_processor.FusedAuraFlowAttnProcessor2_0

< >

( )

Attention processor used typically in processing Aura Flow with fused projections.

CogVideoX

class diffusers.models.attention_processor.CogVideoXAttnProcessor2_0

< >

( )

Processor for implementing scaled dot-product attention for the CogVideoX model. It applies a rotary embedding on query and key vectors, but does not include spatial normalization.

class diffusers.models.attention_processor.FusedCogVideoXAttnProcessor2_0

< >

( )

Processor for implementing scaled dot-product attention for the CogVideoX model. It applies a rotary embedding on query and key vectors, but does not include spatial normalization.

CrossFrameAttnProcessor

class diffusers.pipelines.text_to_video_synthesis.pipeline_text_to_video_zero.CrossFrameAttnProcessor

< >

( batch_size = 2 )

Parameters

  • batch_size — The number that represents actual batch size, other than the frames. For example, calling unet with a single prompt and num_images_per_prompt=1, batch_size should be equal to 2, due to classifier-free guidance.

Cross frame attention processor. Each frame attends the first frame.

Custom Diffusion

class diffusers.models.attention_processor.CustomDiffusionAttnProcessor

< >

( train_kv: bool = True train_q_out: bool = True hidden_size: typing.Optional[int] = None cross_attention_dim: typing.Optional[int] = None out_bias: bool = True dropout: float = 0.0 )

Parameters

  • train_kv (bool, defaults to True) — Whether to newly train the key and value matrices corresponding to the text features.
  • train_q_out (bool, defaults to True) — Whether to newly train query matrices corresponding to the latent image features.
  • hidden_size (int, optional, defaults to None) — The hidden size of the attention layer.
  • cross_attention_dim (int, optional, defaults to None) — The number of channels in the encoder_hidden_states.
  • out_bias (bool, defaults to True) — Whether to include the bias parameter in train_q_out.
  • dropout (float, optional, defaults to 0.0) — The dropout probability to use.

Processor for implementing attention for the Custom Diffusion method.

class diffusers.models.attention_processor.CustomDiffusionAttnProcessor2_0

< >

( train_kv: bool = True train_q_out: bool = True hidden_size: typing.Optional[int] = None cross_attention_dim: typing.Optional[int] = None out_bias: bool = True dropout: float = 0.0 )

Parameters

  • train_kv (bool, defaults to True) — Whether to newly train the key and value matrices corresponding to the text features.
  • train_q_out (bool, defaults to True) — Whether to newly train query matrices corresponding to the latent image features.
  • hidden_size (int, optional, defaults to None) — The hidden size of the attention layer.
  • cross_attention_dim (int, optional, defaults to None) — The number of channels in the encoder_hidden_states.
  • out_bias (bool, defaults to True) — Whether to include the bias parameter in train_q_out.
  • dropout (float, optional, defaults to 0.0) — The dropout probability to use.

Processor for implementing attention for the Custom Diffusion method using PyTorch 2.0’s memory-efficient scaled dot-product attention.

class diffusers.models.attention_processor.CustomDiffusionXFormersAttnProcessor

< >

( train_kv: bool = True train_q_out: bool = False hidden_size: typing.Optional[int] = None cross_attention_dim: typing.Optional[int] = None out_bias: bool = True dropout: float = 0.0 attention_op: typing.Optional[typing.Callable] = None )

Parameters

  • train_kv (bool, defaults to True) — Whether to newly train the key and value matrices corresponding to the text features.
  • train_q_out (bool, defaults to True) — Whether to newly train query matrices corresponding to the latent image features.
  • hidden_size (int, optional, defaults to None) — The hidden size of the attention layer.
  • cross_attention_dim (int, optional, defaults to None) — The number of channels in the encoder_hidden_states.
  • out_bias (bool, defaults to True) — Whether to include the bias parameter in train_q_out.
  • dropout (float, optional, defaults to 0.0) — The dropout probability to use.
  • attention_op (Callable, optional, defaults to None) — The base operator to use as the attention operator. It is recommended to set to None, and allow xFormers to choose the best operator.

Processor for implementing memory efficient attention using xFormers for the Custom Diffusion method.

Flux

class diffusers.models.attention_processor.FluxAttnProcessor2_0

< >

( )

Attention processor used typically in processing the SD3-like self-attention projections.

class diffusers.models.attention_processor.FusedFluxAttnProcessor2_0

< >

( )

Attention processor used typically in processing the SD3-like self-attention projections.

class diffusers.models.attention_processor.FluxSingleAttnProcessor2_0

< >

( )

Processor for implementing scaled dot-product attention (enabled by default if you’re using PyTorch 2.0).

Hunyuan

class diffusers.models.attention_processor.HunyuanAttnProcessor2_0

< >

( )

Processor for implementing scaled dot-product attention (enabled by default if you’re using PyTorch 2.0). This is used in the HunyuanDiT model. It applies a s normalization layer and rotary embedding on query and key vector.

class diffusers.models.attention_processor.FusedHunyuanAttnProcessor2_0

< >

( )

Processor for implementing scaled dot-product attention (enabled by default if you’re using PyTorch 2.0) with fused projection layers. This is used in the HunyuanDiT model. It applies a s normalization layer and rotary embedding on query and key vector.

class diffusers.models.attention_processor.PAGHunyuanAttnProcessor2_0

< >

( )

Processor for implementing scaled dot-product attention (enabled by default if you’re using PyTorch 2.0). This is used in the HunyuanDiT model. It applies a normalization layer and rotary embedding on query and key vector. This variant of the processor employs Pertubed Attention Guidance.

class diffusers.models.attention_processor.PAGCFGHunyuanAttnProcessor2_0

< >

( )

Processor for implementing scaled dot-product attention (enabled by default if you’re using PyTorch 2.0). This is used in the HunyuanDiT model. It applies a normalization layer and rotary embedding on query and key vector. This variant of the processor employs Pertubed Attention Guidance.

IdentitySelfAttnProcessor2_0

class diffusers.models.attention_processor.PAGIdentitySelfAttnProcessor2_0

< >

( )

Processor for implementing PAG using scaled dot-product attention (enabled by default if you’re using PyTorch 2.0). PAG reference: https://arxiv.org/abs/2403.17377

class diffusers.models.attention_processor.PAGCFGIdentitySelfAttnProcessor2_0

< >

( )

Processor for implementing PAG using scaled dot-product attention (enabled by default if you’re using PyTorch 2.0). PAG reference: https://arxiv.org/abs/2403.17377

IP-Adapter

class diffusers.models.attention_processor.IPAdapterAttnProcessor

< >

( hidden_size cross_attention_dim = None num_tokens = (4,) scale = 1.0 )

Parameters

  • hidden_size (int) — The hidden size of the attention layer.
  • cross_attention_dim (int) — The number of channels in the encoder_hidden_states.
  • num_tokens (int, Tuple[int] or List[int], defaults to (4,)) — The context length of the image features.
  • scale (float or Listfloat, defaults to 1.0) — the weight scale of image prompt.

Attention processor for Multiple IP-Adapters.

class diffusers.models.attention_processor.IPAdapterAttnProcessor2_0

< >

( hidden_size cross_attention_dim = None num_tokens = (4,) scale = 1.0 )

Parameters

  • hidden_size (int) — The hidden size of the attention layer.
  • cross_attention_dim (int) — The number of channels in the encoder_hidden_states.
  • num_tokens (int, Tuple[int] or List[int], defaults to (4,)) — The context length of the image features.
  • scale (float or List[float], defaults to 1.0) — the weight scale of image prompt.

Attention processor for IP-Adapter for PyTorch 2.0.

class diffusers.models.attention_processor.SD3IPAdapterJointAttnProcessor2_0

< >

( hidden_size: int ip_hidden_states_dim: int head_dim: int timesteps_emb_dim: int = 1280 scale: float = 0.5 )

Parameters

  • hidden_size (int) — The number of hidden channels.
  • ip_hidden_states_dim (int) — The image feature dimension.
  • head_dim (int) — The number of head channels.
  • timesteps_emb_dim (int, defaults to 1280) — The number of input channels for timestep embedding.
  • scale (float, defaults to 0.5) — IP-Adapter scale.

Attention processor for IP-Adapter used typically in processing the SD3-like self-attention projections, with additional image-based information and timestep embeddings.

JointAttnProcessor2_0

class diffusers.models.attention_processor.JointAttnProcessor2_0

< >

( )

Attention processor used typically in processing the SD3-like self-attention projections.

class diffusers.models.attention_processor.PAGJointAttnProcessor2_0

< >

( )

Attention processor used typically in processing the SD3-like self-attention projections.

class diffusers.models.attention_processor.PAGCFGJointAttnProcessor2_0

< >

( )

Attention processor used typically in processing the SD3-like self-attention projections.

class diffusers.models.attention_processor.FusedJointAttnProcessor2_0

< >

( )

Attention processor used typically in processing the SD3-like self-attention projections.

LoRA

class diffusers.models.attention_processor.LoRAAttnProcessor

< >

( )

Processor for implementing attention with LoRA.

class diffusers.models.attention_processor.LoRAAttnProcessor2_0

< >

( )

Processor for implementing attention with LoRA (enabled by default if you’re using PyTorch 2.0).

class diffusers.models.attention_processor.LoRAAttnAddedKVProcessor

< >

( )

Processor for implementing attention with LoRA with extra learnable key and value matrices for the text encoder.

class diffusers.models.attention_processor.LoRAXFormersAttnProcessor

< >

( )

Processor for implementing attention with LoRA using xFormers.

Lumina-T2X

class diffusers.models.attention_processor.LuminaAttnProcessor2_0

< >

( )

Processor for implementing scaled dot-product attention (enabled by default if you’re using PyTorch 2.0). This is used in the LuminaNextDiT model. It applies a s normalization layer and rotary embedding on query and key vector.

Mochi

class diffusers.models.attention_processor.MochiAttnProcessor2_0

< >

( )

Attention processor used in Mochi.

class diffusers.models.attention_processor.MochiVaeAttnProcessor2_0

< >

( )

Attention processor used in Mochi VAE.

Sana

class diffusers.models.attention_processor.SanaLinearAttnProcessor2_0

< >

( )

Processor for implementing scaled dot-product linear attention.

class diffusers.models.attention_processor.SanaMultiscaleAttnProcessor2_0

< >

( )

Processor for implementing multiscale quadratic attention.

class diffusers.models.attention_processor.PAGCFGSanaLinearAttnProcessor2_0

< >

( )

Processor for implementing scaled dot-product linear attention.

class diffusers.models.attention_processor.PAGIdentitySanaLinearAttnProcessor2_0

< >

( )

Processor for implementing scaled dot-product linear attention.

Stable Audio

class diffusers.models.attention_processor.StableAudioAttnProcessor2_0

< >

( )

Processor for implementing scaled dot-product attention (enabled by default if you’re using PyTorch 2.0). This is used in the Stable Audio model. It applies rotary embedding on query and key vector, and allows MHA, GQA or MQA.

SlicedAttnProcessor

class diffusers.models.attention_processor.SlicedAttnProcessor

< >

( slice_size: int )

Parameters

  • slice_size (int, optional) — The number of steps to compute attention. Uses as many slices as attention_head_dim // slice_size, and attention_head_dim must be a multiple of the slice_size.

Processor for implementing sliced attention.

class diffusers.models.attention_processor.SlicedAttnAddedKVProcessor

< >

( slice_size )

Parameters

  • slice_size (int, optional) — The number of steps to compute attention. Uses as many slices as attention_head_dim // slice_size, and attention_head_dim must be a multiple of the slice_size.

Processor for implementing sliced attention with extra learnable key and value matrices for the text encoder.

XFormersAttnProcessor

class diffusers.models.attention_processor.XFormersAttnProcessor

< >

( attention_op: typing.Optional[typing.Callable] = None )

Parameters

  • attention_op (Callable, optional, defaults to None) — The base operator to use as the attention operator. It is recommended to set to None, and allow xFormers to choose the best operator.

Processor for implementing memory efficient attention using xFormers.

class diffusers.models.attention_processor.XFormersAttnAddedKVProcessor

< >

( attention_op: typing.Optional[typing.Callable] = None )

Parameters

  • attention_op (Callable, optional, defaults to None) — The base operator to use as the attention operator. It is recommended to set to None, and allow xFormers to choose the best operator.

Processor for implementing memory efficient attention using xFormers.

XLAFlashAttnProcessor2_0

class diffusers.models.attention_processor.XLAFlashAttnProcessor2_0

< >

( partition_spec: typing.Optional[typing.Tuple[typing.Optional[str], ...]] = None )

Processor for implementing scaled dot-product attention with pallas flash attention kernel if using torch_xla.

< > Update on GitHub

←Overview Custom activation functions→