> For the complete documentation index, see [llms.txt](https://boinc-ai.gitbook.io/diffusers/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://boinc-ai.gitbook.io/diffusers/api/models/transformer-temporal.md).

# Transformer Temporal

## Transformer Temporal

A Transformer model for video-like data.

### TransformerTemporalModel

#### class diffusers.models.TransformerTemporalModel

[\<source>](https://github.com/huggingface/diffusers/blob/v0.21.0/src/diffusers/models/transformer_temporal.py#L39)

( num\_attention\_heads: int = 16attention\_head\_dim: int = 88in\_channels: typing.Optional\[int] = Noneout\_channels: typing.Optional\[int] = Nonenum\_layers: int = 1dropout: float = 0.0norm\_num\_groups: int = 32cross\_attention\_dim: typing.Optional\[int] = Noneattention\_bias: bool = Falsesample\_size: typing.Optional\[int] = Noneactivation\_fn: str = 'geglu'norm\_elementwise\_affine: bool = Truedouble\_self\_attention: bool = True )

Parameters

* **num\_attention\_heads** (`int`, *optional*, defaults to 16) — The number of heads to use for multi-head attention.
* **attention\_head\_dim** (`int`, *optional*, defaults to 88) — The number of channels in each head.
* **in\_channels** (`int`, *optional*) — The number of channels in the input and output (specify if the input is **continuous**).
* **num\_layers** (`int`, *optional*, defaults to 1) — The number of layers of Transformer blocks to use.
* **dropout** (`float`, *optional*, defaults to 0.0) — The dropout probability to use.
* **cross\_attention\_dim** (`int`, *optional*) — The number of `encoder_hidden_states` dimensions to use.
* **sample\_size** (`int`, *optional*) — The width of the latent images (specify if the input is **discrete**). This is fixed during training since it is used to learn a number of position embeddings.
* **activation\_fn** (`str`, *optional*, defaults to `"geglu"`) — Activation function to use in feed-forward.
* **attention\_bias** (`bool`, *optional*) — Configure if the `TransformerBlock` attention should contain a bias parameter.
* **double\_self\_attention** (`bool`, *optional*) — Configure if each `TransformerBlock` should contain two self-attention layers.

A Transformer model for video-like data.

**forward**

[\<source>](https://github.com/huggingface/diffusers/blob/v0.21.0/src/diffusers/models/transformer_temporal.py#L107)

( hidden\_statesencoder\_hidden\_states = Nonetimestep = Noneclass\_labels = Nonenum\_frames = 1cross\_attention\_kwargs = Nonereturn\_dict: bool = True ) → [TransformerTemporalModelOutput](https://huggingface.co/docs/diffusers/v0.21.0/en/api/models/transformer_temporal#diffusers.models.transformer_temporal.TransformerTemporalModelOutput) or `tuple`

Parameters

* **hidden\_states** (`torch.LongTensor` of shape `(batch size, num latent pixels)` if discrete, `torch.FloatTensor` of shape `(batch size, channel, height, width)` if continuous) — Input hidden\_states.
* **encoder\_hidden\_states** ( `torch.LongTensor` of shape `(batch size, encoder_hidden_states dim)`, *optional*) — Conditional embeddings for cross attention layer. If not given, cross-attention defaults to self-attention.
* **timestep** ( `torch.long`, *optional*) — Used to indicate denoising step. Optional timestep to be applied as an embedding in `AdaLayerNorm`.
* **class\_labels** ( `torch.LongTensor` of shape `(batch size, num classes)`, *optional*) — Used to indicate class labels conditioning. Optional class labels to be applied as an embedding in `AdaLayerZeroNorm`.
* **return\_dict** (`bool`, *optional*, defaults to `True`) — Whether or not to return a [UNet2DConditionOutput](https://huggingface.co/docs/diffusers/v0.21.0/en/api/models/unet2d-cond#diffusers.models.unet_2d_condition.UNet2DConditionOutput) instead of a plain tuple.

Returns

[TransformerTemporalModelOutput](https://huggingface.co/docs/diffusers/v0.21.0/en/api/models/transformer_temporal#diffusers.models.transformer_temporal.TransformerTemporalModelOutput) or `tuple`

If `return_dict` is True, an [TransformerTemporalModelOutput](https://huggingface.co/docs/diffusers/v0.21.0/en/api/models/transformer_temporal#diffusers.models.transformer_temporal.TransformerTemporalModelOutput) is returned, otherwise a `tuple` where the first element is the sample tensor.

The `TransformerTemporal` forward method.

### TransformerTemporalModelOutput

#### class diffusers.models.transformer\_temporal.TransformerTemporalModelOutput

[\<source>](https://github.com/huggingface/diffusers/blob/v0.21.0/src/diffusers/models/transformer_temporal.py#L27)

( sample: FloatTensor )

Parameters

* **sample** (`torch.FloatTensor` of shape `(batch_size x num_frames, num_channels, height, width)`) — The hidden states output conditioned on `encoder_hidden_states` input.

The output of `TransformerTemporalModel`.