Perceiver IO Decoder

class vformer.decoder.perceiver_io.PerceiverIODecoder(dim=32, latent_dim=512, queries_dim=32, num_cross_heads=1, cross_head_dim=64, logits_dim=None, decoder_ff=False)[source]

Implementation of the Perceiver IO Decoder

  • dim (int) – Size of sequence to be encoded

  • latent_dim (int) – Dimension of latent array

  • queries_dim (int) – Dimension of queries array

  • num_latents (int) – Number of latent arrays

  • num_cross_heads (int) – Number of heads for cross attention

  • cross_head_dim (int) – Dimension of cross attention head

  • logits_dim (int, optional) – Dimension of output logits

  • decoder_ff (bool) – Whether to include a feed forward layer for the decoder attention block

forward(x, mask=None, queries=None)[source]

Defines the computation performed at every call.

Should be overridden by all subclasses.


Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.