Closed gopi-erabati closed 9 months ago
Is the recurrent representation of retentive network only valid for causal nature of input as such in language tasks ? If the input nature is without causal property, will the recurrent representation hold true ?
We can also apply autoregressive modeling for image data, such as iGPT and DALL-E v1, although the nature of images is not causal.
Is the recurrent representation of retentive network only valid for causal nature of input as such in language tasks ? If the input nature is without causal property, will the recurrent representation hold true ?