Class TransformerTemporalModel

animation/modules/attention.py:27–185

A Transformer model for video-like data. 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_channel

class TransformerTemporalModel(ModelMixin, ConfigMixin):
    """
    A Transformer model for video-like data.

    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.
        attention_bias (`bool`, *optional*):
            Configure if the `TransformerBlock` attention should contain a bias parameter.
        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. See `diffusers.models.activations.get_activation` for supported
            activation functions.
        norm_elementwise_affine (`bool`, *optional*):
            Configure if the `TransformerBlock` should use learnable elementwise affine parameters for normalization.
        double_self_attention (`bool`, *optional*):
            Configure if each `TransformerBlock` should contain two self-attention layers.
        positional_embeddings: (`str`, *optional*):
            The type of positional embeddings to apply to the sequence input before passing use.
        num_positional_embeddings: (`int`, *optional*):
            The maximum length of the sequence over which to apply positional embeddings.
    """

    @register_to_config
    def __init__(
            self,
            num_attention_heads: int = 16,
            attention_head_dim: int = 88,
            in_channels: Optional[int] = None,
            out_channels: Optional[int] = None,
            num_layers: int = 1,
            dropout: float = 0.0,
            norm_num_groups: int = 32,
            cross_attention_dim: Optional[int] = None,
            attention_bias: bool = False,
            sample_size: Optional[int] = None,
            activation_fn: str = "geglu",
            norm_elementwise_affine: bool = True,
            double_self_attention: bool = True,
            positional_embeddings: Optional[str] = None,
            num_positional_embeddings: Optional[int] = None,
    ):
        super().__init__()
        self.num_attention_heads = num_attention_heads
        self.attention_head_dim = attention_head_dim
        inner_dim = num_attention_heads * attention_head_dim

        self.in_channels = in_channels

        self.norm = torch.nn.GroupNorm(num_groups=norm_num_groups, num_channels=in_channels, eps=1e-6, affine=True)
        self.proj_in = nn.Linear(in_channels, inner_dim)