1.5.0

switch to causal convolution for conv-like attention using unfold

1.4.2

1.4.2

1.4.1

remove g-mlps

1.4.0

1.4.0 for improved inference speeds, thanks to @borzunov

1.2.2

make sure generate_text works when shift token feature is turned on

1.2.1

add super-conditioning feature, idea originated from @crowsonkb

1.2.0

remove some dead code, text mask is not needed when training DALL-E

1.1.8

remove softmax and torch multinomial when sampling

1.1.7

stable softmax when generating images

1.1.6

start storing version of dalle-pytorch alongside model weights - also…